# Coral Edge TPU on a Raspberry Pi with Ultralytics YOLOv8. 🚀
<palign="center">
<imgwidth="800"src="https://images.ctfassets.net/2lpsze4g694w/5XK2dV0w55U0TefijPli1H/bf0d119d77faef9a5d2cc0dad2aa4b42/Edge-TPU-USB-Accelerator-and-Pi.jpg?w=800"alt="Raspberry Pi single board computer with USB Edge TPU accelerator">
</p>
## What is a Coral Edge TPU?
The Coral USB Accelerator is a compact device that adds an Edge TPU coprocessor to your system. It enables low-power, high-performance ML inferencing for TensorFlow Lite models. Read more at the [Coral Edge TPU home page](https://coral.ai/products/accelerator).
## Why?
Many people want to run their models on an embedded or mobile device such as a Raspberry Pi, since they are very power efficient and can be used in many different applications. However, the inference performance on these devices is usually poor even when using formats like `onnx` or `openvino`. The Coral Edge TPU is a great solution to this problem, since it can be used with a Raspberry Pi and accelerate inference performance greatly.
## Why this guide?
The [current guide](https://coral.ai/docs/accelerator/get-started/) by Coral on how to use the Edge TPU with a Raspberry Pi is outdated, and the current Coral Edge TPU runtime builds do not work with the current TensorFlow Lite runtime versions anymore. In addition to that, Google seems to have completely abandoned the Coral project, and there have not been any updates between 2021 and 2024. This guide will show you how to get the Edge TPU working with the latest versions of the TensorFlow Lite runtime and an updated Coral Edge TPU runtime on a Raspberry Pi single board computer (SBC).
## Prerequisites
- [Raspberry Pi 4B](https://www.raspberrypi.com/products/raspberry-pi-4-model-b/) (2GB or more recommended) or [Raspberry Pi 5](https://www.raspberrypi.com/products/raspberry-pi-5/) (Recommended)
- [Raspberry Pi OS](https://www.raspberrypi.com/software/) Bullseye/Bookworm (64-bit) with desktop (Recommended)
- [Coral USB Accelerator](https://coral.ai/products/accelerator/)
- A non-ARM based platform for exporting an Ultralytics PyTorch model
## Installation
This guide assumes that you already have a working Raspberry Pi OS install and have installed `ultralytics` and all dependencies. To get `ultralytics` installed, visit the [quickstart guide](../quickstart.md) to get setup before continuing here.
### Installing the Edge TPU runtime
First, we need to install the Edge TPU runtime. There are many different versions available, so you need to choose the version for your device and operating system.
| Raspberry Pi OS | High frequency mode | Version to download |
[Download the latest version](https://github.com/feranick/libedgetpu/releases).
After downloading the file, you can install it with the following command:
```bash
sudo dpkg -i path/to/file
```
After installing the runtime, you need to plug in your Coral Edge TPU into a USB 3.0 port on your Raspberry Pi. This is because, according to the official guide, a new `udev` rule needs to take effect after installation.
???+ warning "Important"
If you already have the Coral Edge TPU runtime installed, uninstall it using the following command.
```bash
# If you installed the standard version
sudo apt remove libedgetpu1-std
# If you installed the high frequency version
sudo apt remove libedgetpu1-max
```
## Exporting your model to a Edge TPU compatible model
To use the Edge TPU, you need to convert your model into a compatible format. It is recommended that you run export on Google Colab, x86_64 Linux machine, using the official [Ultralytics Docker container](docker-quickstart.md), or using [Ultralytics HUB](../hub/quickstart.md), since the Edge TPU compiler is not available on ARM. See the [Export Mode](../modes/export.md) for the available arguments.
!!! Exporting the model
=== "Python"
```python
from ultralytics import YOLO
# Load a model
model = YOLO('path/to/model.pt') # Load a official model or custom model
# Export the model
model.export(format='edgetpu')
```
=== "CLI"
```bash
yolo export model=path/to/model.pt format=edgetpu # Export a official model or custom model
```
The exported model will be saved in the `<model_name>_saved_model/` folder with the name `<model_name>_full_integer_quant_edgetpu.tflite`.
## Running the model
After exporting your model, you can run inference with it using the following code:
!!! Running the model
=== "Python"
```python
from ultralytics import YOLO
# Load a model
model = YOLO('path/to/model.tflite') # Load a official model or custom model
# Run Prediction
model.predict("path/to/source.png")
```
For a more detailed guide on how to use the Predict Mode for inference, visit the [Prediction Mode](../modes/predict.md) page.
???+ warning "Important"
You should run the model using `tflite-runtime` and not `tensorflow`.
If `tensorflow` is installed, uninstall tensorflow with the following command:
```bash
pip uninstall tensorflow tensorflow-aarch64
```
Then install/update `tflite-runtime`:
```
pip install -U tflite-runtime
```
If you want a `tflite-runtime` wheel for `tensorflow` 2.15.0 download it from [here](https://github.com/feranick/TFlite-builds/releases) and install it using `pip` or your package manager of choice.
@ -38,6 +38,7 @@ Here's a compilation of in-depth guides to help you master different aspects of
- [Triton Inference Server Integration](triton-inference-server.md) 🚀 NEW: Dive into the integration of Ultralytics YOLOv8 with NVIDIA's Triton Inference Server for scalable and efficient deep learning inference deployments.
- [YOLO Thread-Safe Inference](yolo-thread-safe-inference.md) 🚀 NEW: Guidelines for performing inference with YOLO models in a thread-safe manner. Learn the importance of thread safety and best practices to prevent race conditions and ensure consistent predictions.
- [Isolating Segmentation Objects](isolating-segmentation-objects.md) 🚀 NEW: Step-by-step recipe and explanation on how to extract and/or isolate objects from images using Ultralytics Segmentation.
- [Edge TPU on Raspberry Pi](coral-edge-tpu-on-raspberry-pi.md): [Google Edge TPU](https://coral.ai/products/accelerator) accelerates YOLO inference on [Raspberry Pi](https://www.raspberrypi.com/).
| `source` | `str` | `'ultralytics/assets'` | Specifies the data source for inference. Can be an image path, video file, directory, URL, or device ID for live feeds. Supports a wide range of formats and sources, enabling flexible application across different types of input. |
| `conf` | `float` | `0.25` | Sets the minimum confidence threshold for detections. Objects detected with confidence below this threshold will be disregarded. Adjusting this value can help reduce false positives. |
| `iou` | `float` | `0.7` | Intersection Over Union (IoU) threshold for Non-Maximum Suppression (NMS). Higher values result in fewer detections by eliminating overlapping boxes, useful for reducing duplicates. |
| `imgsz` | `int or tuple` | `640` | Defines the image size for inference. Can be a single integer `640` for square resizing or a (height, width) tuple. Proper sizing can improve detection accuracy and processing speed. |
| `half` | `bool` | `False` | Enables half-precision (FP16) inference, which can speed up model inference on supported GPUs with minimal impact on accuracy. |
| `device` | `str` | `None` | Specifies the device for inference (e.g., `cpu`, `cuda:0` or `0`). Allows users to select between CPU, a specific GPU, or other compute devices for model execution. |
| `max_det` | `int` | `300` | Maximum number of detections allowed per image. Limits the total number of objects the model can detect in a single inference, preventing excessive outputs in dense scenes. |
| `vid_stride` | `int` | `1` | Frame stride for video inputs. Allows skipping frames in videos to speed up processing at the cost of temporal resolution. A value of 1 processes every frame, higher values skip frames. |
| `stream_buffer` | `bool` | `False` | Determines if all frames should be buffered when processing video streams (`True`), or if the model should return the most recent frame (`False`). Useful for real-time applications. |
| `visualize` | `bool` | `False` | Activates visualization of model features during inference, providing insights into what the model is "seeing". Useful for debugging and model interpretation. |
| `augment` | `bool` | `False` | Enables test-time augmentation (TTA) for predictions, potentially improving detection robustness at the cost of inference speed. |
| `agnostic_nms` | `bool` | `False` | Enables class-agnostic Non-Maximum Suppression(NMS),whichmergesoverlappingboxesofdifferent classes. Useful in multi-class detection scenarios where class overlap is common. |
| `classes` | `list[int]` | `None` | Filters predictions to a set of class IDs. Only detections belonging to the specified classes will be returned. Useful for focusing on relevant objects in multi-class detection tasks. |
| `retina_masks` | `bool` | `False` | Uses high-resolution segmentation masks if available in the model. This can enhance mask quality for segmentation tasks, providing finer detail. |
| `embed` | `list[int]` | `None` | Specifies the layers from which to extract feature vectors or embeddings. Useful for downstream tasks like clustering or similarity search. |
| `show` | `bool` | `False` | If `True`, displays the annotated images or videos in a window. Useful for immediate visual feedback during development or testing. |
| `save` | `bool` | `False` | Enables saving of the annotated images or videos to file. Useful for documentation, further analysis, or sharing results. |
| `save_frames` | `bool` | `False` | When processing videos, saves individual frames as images. Useful for extracting specific frames or for detailed frame-by-frame analysis. |
| `save_txt` | `bool` | `False` | Saves detection results in a text file, following the format `[class] [x_center] [y_center] [width] [height] [confidence]`. Useful for integration with other analysis tools. |
| `save_conf` | `bool` | `False` | Includes confidence scores in the saved text files. Enhances the detail available for post-processing and analysis. |
| `save_crop` | `bool` | `False` | Saves cropped images of detections. Useful for dataset augmentation, analysis, or creating focused datasets for specific objects. |
| `show_labels` | `bool` | `True` | Displays labels for each detection in the visual output. Provides immediate understanding of detected objects. |
| `show_conf` | `bool` | `True` | Displays the confidence score for each detection alongside the label. Gives insight into the model's certainty for each detection. |
| `show_boxes` | `bool` | `True` | Draws bounding boxes around detected objects. Essential for visual identification and location of objects in images or video frames. |
| `line_width` | `None or int` | `None` | Specifies the line width of bounding boxes. If `None`, the line width is automatically adjusted based on the image size. Provides visual customization for clarity. |
@ -175,56 +175,56 @@ Remember that checkpoints are saved at the end of every epoch by default, or at
Training settings for YOLO models refer to the various hyperparameters and configurations used to train the model on a dataset. These settings can affect the model's performance, speed, and accuracy. Some common YOLO training settings include the batch size, learning rate, momentum, and weight decay. Other factors that may affect the training process include the choice of optimizer, the choice of loss function, and the size and composition of the training dataset. It is important to carefully tune and experiment with these settings to achieve the best possible performance for a given task.
| `model` | `None` | Specifies the model file for training. Accepts a path to either a `.pt` pretrained model or a `.yaml` configuration file. Essential for defining the model structure or initializing weights. |
| `data` | `None` | Path to the dataset configuration file (e.g., `coco128.yaml`). This file contains dataset-specific parameters, including paths to training and validation data, class names, and number of classes. |
| `epochs` | `100` | Total number of training epochs. Each epoch represents a full pass over the entire dataset. Adjusting this value can affect training duration and model performance. |
| `time` | `None` | Maximum training time in hours. If set, this overrides the `epochs` argument, allowing training to automatically stop after the specified duration. Useful for time-constrained training scenarios. |
| `patience` | `50` | Number of epochs to wait without improvement in validation metrics before early stopping the training. Helps prevent overfitting by stopping training when performance plateaus. |
| `batch` | `16` | Batch size for training, indicating how many images are processed before the model's internal parameters are updated. AutoBatch (`batch=-1`) dynamically adjusts the batch size based on GPU memory availability. |
| `imgsz` | `640` | Target image size for training. All images are resized to this dimension before being fed into the model. Affects model accuracy and computational complexity. |
| `save` | `True` | Enables saving of training checkpoints and final model weights. Useful for resuming training or model deployment. |
| `save_period` | `-1` | Frequency of saving model checkpoints, specified in epochs. A value of -1 disables this feature. Useful for saving interim models during long training sessions. |
| `cache` | `False` | Enables caching of dataset images in memory (`True`/`ram`), on disk (`disk`), or disables it (`False`). Improves training speed by reducing disk I/O at the cost of increased memory usage. |
| `device` | `None` | Specifies the computational device(s) for training: a single GPU (`device=0`), multiple GPUs (`device=0,1`), CPU (`device=cpu`), or MPS for Apple silicon (`device=mps`). |
| `workers` | `8` | Number of worker threads for data loading (per `RANK` if Multi-GPU training). Influences the speed of data preprocessing and feeding into the model, especially useful in multi-GPU setups. |
| `project` | `None` | Name of the project directory where training outputs are saved. Allows for organized storage of different experiments. |
| `name` | `None` | Name of the training run. Used for creating a subdirectory within the project folder, where training logs and outputs are stored. |
| `exist_ok` | `False` | If True, allows overwriting of an existing project/name directory. Useful for iterative experimentation without needing to manually clear previous outputs. |
| `pretrained` | `True` | Determines whether to start training from a pretrained model. Can be a boolean value or a string path to a specific model from which to load weights. Enhances training efficiency and model performance. |
| `optimizer` | `'auto'` | Choice of optimizer for training. Options include `SGD`, `Adam`, `AdamW`, `NAdam`, `RAdam`, `RMSProp` etc., or `auto` for automatic selection based on model configuration. Affects convergence speed and stability. |
| `verbose` | `False` | Enables verbose output during training, providing detailed logs and progress updates. Useful for debugging and closely monitoring the training process. |
| `seed` | `0` | Sets the random seed for training, ensuring reproducibility of results across runs with the same configurations. |
| `deterministic` | `True` | Forces deterministic algorithm use, ensuring reproducibility but may affect performance and speed due to the restriction on non-deterministic algorithms. |
| `single_cls` | `False` | Treats all classes in multi-class datasets as a single class during training. Useful for binary classification tasks or when focusing on object presence rather than classification. |
| `rect` | `False` | Enables rectangular training, optimizing batch composition for minimal padding. Can improve efficiency and speed but may affect model accuracy. |
| `cos_lr` | `False` | Utilizes a cosine learning rate scheduler, adjusting the learning rate following a cosine curve over epochs. Helps in managing learning rate for better convergence. |
| `close_mosaic` | `10` | Disables mosaic data augmentation in the last N epochs to stabilize training before completion. Setting to 0 disables this feature. |
| `resume` | `False` | Resumes training from the last saved checkpoint. Automatically loads model weights, optimizer state, and epoch count, continuing training seamlessly. |
| `amp` | `True` | Enables Automatic Mixed Precision (AMP) training, reducing memory usage and possibly speeding up training with minimal impact on accuracy. |
| `fraction` | `1.0` | Specifies the fraction of the dataset to use for training. Allows for training on a subset of the full dataset, useful for experiments or when resources are limited. |
| `profile` | `False` | Enables profiling of ONNX and TensorRT speeds during training, useful for optimizing model deployment. |
| `freeze` | `None` | Freezes the first N layers of the model or specified layers by index, reducing the number of trainable parameters. Useful for fine-tuning or transfer learning. |
| `lr0` | `0.01` | Initial learning rate (i.e. `SGD=1E-2`, `Adam=1E-3`) . Adjusting this value is crucial for the optimization process, influencing how rapidly model weights are updated. |
| `lrf` | `0.01` | Final learning rate as a fraction of the initial rate = (`lr0 * lrf`), used in conjunction with schedulers to adjust the learning rate over time. |
| `momentum` | `0.937` | Momentum factor for SGD or beta1 for Adam optimizers, influencing the incorporation of past gradients in the current update. |
| `weight_decay` | `0.0005` | L2 regularization term, penalizing large weights to prevent overfitting. |
| `warmup_epochs` | `3.0` | Number of epochs for learning rate warmup, gradually increasing the learning rate from a low value to the initial learning rate to stabilize training early on. |
| `warmup_momentum` | `0.8` | Initial momentum for warmup phase, gradually adjusting to the set momentum over the warmup period. |
| `warmup_bias_lr` | `0.1` | Learning rate for bias parameters during the warmup phase, helping stabilize model training in the initial epochs. |
| `box` | `7.5` | Weight of the box loss component in the loss function, influencing how much emphasis is placed on accurately predicting bounding box coordinates. |
| `cls` | `0.5` | Weight of the classification loss in the total loss function, affecting the importance of correct class prediction relative to other components. |
| `dfl` | `1.5` | Weight of the distribution focal loss,used in certain YOLO versions for fine-grained classification. |
| `pose` | `12.0` | Weight of the pose loss in models trained for pose estimation, influencing the emphasis on accurately predicting pose keypoints. |
| `kobj` | `2.0` | Weight of the keypoint objectness loss in pose estimation models, balancing detection confidence with pose accuracy. |
| `label_smoothing` | `0.0` | Applies label smoothing, softening hard labels to a mix of the target label and a uniform distribution over labels, can improve generalization. |
| `nbs` | `64` | Nominal batch size for normalization of loss. |
| `overlap_mask` | `True` | Determines whether segmentation masks should overlap during training, applicable in instance segmentation tasks. |
| `mask_ratio` | `4` | Downsample ratio for segmentation masks, affecting the resolution of masks used during training. |
| `dropout` | `0.0` | Dropout rate for regularization in classification tasks, preventing overfitting by randomly omitting units during training. |
| `val` | `True` | Enables validation during training, allowing for periodic evaluation of model performance on a separate dataset. |
| `plots` | `False` | Generates and saves plots of training and validation metrics, as well as prediction examples, providing visual insights into model performance and learning progression. |
@ -83,56 +83,56 @@ YOLO models can be used in different modes depending on the specific problem you
The training settings for YOLO models encompass various hyperparameters and configurations used during the training process. These settings influence the model's performance, speed, and accuracy. Key training settings include batch size, learning rate, momentum, and weight decay. Additionally, the choice of optimizer, loss function, and training dataset composition can impact the training process. Careful tuning and experimentation with these settings are crucial for optimizing performance.
| `model` | `None` | Specifies the model file for training. Accepts a path to either a `.pt` pretrained model or a `.yaml` configuration file. Essential for defining the model structure or initializing weights. |
| `data` | `None` | Path to the dataset configuration file (e.g., `coco128.yaml`). This file contains dataset-specific parameters, including paths to training and validation data, class names, and number of classes. |
| `epochs` | `100` | Total number of training epochs. Each epoch represents a full pass over the entire dataset. Adjusting this value can affect training duration and model performance. |
| `time` | `None` | Maximum training time in hours. If set, this overrides the `epochs` argument, allowing training to automatically stop after the specified duration. Useful for time-constrained training scenarios. |
| `patience` | `50` | Number of epochs to wait without improvement in validation metrics before early stopping the training. Helps prevent overfitting by stopping training when performance plateaus. |
| `batch` | `16` | Batch size for training, indicating how many images are processed before the model's internal parameters are updated. AutoBatch (`batch=-1`) dynamically adjusts the batch size based on GPU memory availability. |
| `imgsz` | `640` | Target image size for training. All images are resized to this dimension before being fed into the model. Affects model accuracy and computational complexity. |
| `save` | `True` | Enables saving of training checkpoints and final model weights. Useful for resuming training or model deployment. |
| `save_period` | `-1` | Frequency of saving model checkpoints, specified in epochs. A value of -1 disables this feature. Useful for saving interim models during long training sessions. |
| `cache` | `False` | Enables caching of dataset images in memory (`True`/`ram`), on disk (`disk`), or disables it (`False`). Improves training speed by reducing disk I/O at the cost of increased memory usage. |
| `device` | `None` | Specifies the computational device(s) for training: a single GPU (`device=0`), multiple GPUs (`device=0,1`), CPU (`device=cpu`), or MPS for Apple silicon (`device=mps`). |
| `workers` | `8` | Number of worker threads for data loading (per `RANK` if Multi-GPU training). Influences the speed of data preprocessing and feeding into the model, especially useful in multi-GPU setups. |
| `project` | `None` | Name of the project directory where training outputs are saved. Allows for organized storage of different experiments. |
| `name` | `None` | Name of the training run. Used for creating a subdirectory within the project folder, where training logs and outputs are stored. |
| `exist_ok` | `False` | If True, allows overwriting of an existing project/name directory. Useful for iterative experimentation without needing to manually clear previous outputs. |
| `pretrained` | `True` | Determines whether to start training from a pretrained model. Can be a boolean value or a string path to a specific model from which to load weights. Enhances training efficiency and model performance. |
| `optimizer` | `'auto'` | Choice of optimizer for training. Options include `SGD`, `Adam`, `AdamW`, `NAdam`, `RAdam`, `RMSProp` etc., or `auto` for automatic selection based on model configuration. Affects convergence speed and stability. |
| `verbose` | `False` | Enables verbose output during training, providing detailed logs and progress updates. Useful for debugging and closely monitoring the training process. |
| `seed` | `0` | Sets the random seed for training, ensuring reproducibility of results across runs with the same configurations. |
| `deterministic` | `True` | Forces deterministic algorithm use, ensuring reproducibility but may affect performance and speed due to the restriction on non-deterministic algorithms. |
| `single_cls` | `False` | Treats all classes in multi-class datasets as a single class during training. Useful for binary classification tasks or when focusing on object presence rather than classification. |
| `rect` | `False` | Enables rectangular training, optimizing batch composition for minimal padding. Can improve efficiency and speed but may affect model accuracy. |
| `cos_lr` | `False` | Utilizes a cosine learning rate scheduler, adjusting the learning rate following a cosine curve over epochs. Helps in managing learning rate for better convergence. |
| `close_mosaic` | `10` | Disables mosaic data augmentation in the last N epochs to stabilize training before completion. Setting to 0 disables this feature. |
| `resume` | `False` | Resumes training from the last saved checkpoint. Automatically loads model weights, optimizer state, and epoch count, continuing training seamlessly. |
| `amp` | `True` | Enables Automatic Mixed Precision (AMP) training, reducing memory usage and possibly speeding up training with minimal impact on accuracy. |
| `fraction` | `1.0` | Specifies the fraction of the dataset to use for training. Allows for training on a subset of the full dataset, useful for experiments or when resources are limited. |
| `profile` | `False` | Enables profiling of ONNX and TensorRT speeds during training, useful for optimizing model deployment. |
| `freeze` | `None` | Freezes the first N layers of the model or specified layers by index, reducing the number of trainable parameters. Useful for fine-tuning or transfer learning. |
| `lr0` | `0.01` | Initial learning rate (i.e. `SGD=1E-2`, `Adam=1E-3`) . Adjusting this value is crucial for the optimization process, influencing how rapidly model weights are updated. |
| `lrf` | `0.01` | Final learning rate as a fraction of the initial rate = (`lr0 * lrf`), used in conjunction with schedulers to adjust the learning rate over time. |
| `momentum` | `0.937` | Momentum factor for SGD or beta1 for Adam optimizers, influencing the incorporation of past gradients in the current update. |
| `weight_decay` | `0.0005` | L2 regularization term, penalizing large weights to prevent overfitting. |
| `warmup_epochs` | `3.0` | Number of epochs for learning rate warmup, gradually increasing the learning rate from a low value to the initial learning rate to stabilize training early on. |
| `warmup_momentum` | `0.8` | Initial momentum for warmup phase, gradually adjusting to the set momentum over the warmup period. |
| `warmup_bias_lr` | `0.1` | Learning rate for bias parameters during the warmup phase, helping stabilize model training in the initial epochs. |
| `box` | `7.5` | Weight of the box loss component in the loss function, influencing how much emphasis is placed on accurately predicting bounding box coordinates. |
| `cls` | `0.5` | Weight of the classification loss in the total loss function, affecting the importance of correct class prediction relative to other components. |
| `dfl` | `1.5` | Weight of the distribution focal loss,used in certain YOLO versions for fine-grained classification. |
| `pose` | `12.0` | Weight of the pose loss in models trained for pose estimation, influencing the emphasis on accurately predicting pose keypoints. |
| `kobj` | `2.0` | Weight of the keypoint objectness loss in pose estimation models, balancing detection confidence with pose accuracy. |
| `label_smoothing` | `0.0` | Applies label smoothing, softening hard labels to a mix of the target label and a uniform distribution over labels, can improve generalization. |
| `nbs` | `64` | Nominal batch size for normalization of loss. |
| `overlap_mask` | `True` | Determines whether segmentation masks should overlap during training, applicable in instance segmentation tasks. |
| `mask_ratio` | `4` | Downsample ratio for segmentation masks, affecting the resolution of masks used during training. |
| `dropout` | `0.0` | Dropout rate for regularization in classification tasks, preventing overfitting by randomly omitting units during training. |
| `val` | `True` | Enables validation during training, allowing for periodic evaluation of model performance on a separate dataset. |
| `plots` | `False` | Generates and saves plots of training and validation metrics, as well as prediction examples, providing visual insights into model performance and learning progression. |
[Train Guide](../modes/train.md){ .md-button }
@ -142,38 +142,38 @@ The prediction settings for YOLO models encompass a range of hyperparameters and
| `source` | `str` | `'ultralytics/assets'` | Specifies the data source for inference. Can be an image path, video file, directory, URL, or device ID for live feeds. Supports a wide range of formats and sources, enabling flexible application across different types of input. |
| `conf` | `float` | `0.25` | Sets the minimum confidence threshold for detections. Objects detected with confidence below this threshold will be disregarded. Adjusting this value can help reduce false positives. |
| `iou` | `float` | `0.7` | Intersection Over Union (IoU) threshold for Non-Maximum Suppression (NMS). Higher values result in fewer detections by eliminating overlapping boxes, useful for reducing duplicates. |
| `imgsz` | `int or tuple` | `640` | Defines the image size for inference. Can be a single integer `640` for square resizing or a (height, width) tuple. Proper sizing can improve detection accuracy and processing speed. |
| `half` | `bool` | `False` | Enables half-precision (FP16) inference, which can speed up model inference on supported GPUs with minimal impact on accuracy. |
| `device` | `str` | `None` | Specifies the device for inference (e.g., `cpu`, `cuda:0` or `0`). Allows users to select between CPU, a specific GPU, or other compute devices for model execution. |
| `max_det` | `int` | `300` | Maximum number of detections allowed per image. Limits the total number of objects the model can detect in a single inference, preventing excessive outputs in dense scenes. |
| `vid_stride` | `int` | `1` | Frame stride for video inputs. Allows skipping frames in videos to speed up processing at the cost of temporal resolution. A value of 1 processes every frame, higher values skip frames. |
| `stream_buffer` | `bool` | `False` | Determines if all frames should be buffered when processing video streams (`True`), or if the model should return the most recent frame (`False`). Useful for real-time applications. |
| `visualize` | `bool` | `False` | Activates visualization of model features during inference, providing insights into what the model is "seeing". Useful for debugging and model interpretation. |
| `augment` | `bool` | `False` | Enables test-time augmentation (TTA) for predictions, potentially improving detection robustness at the cost of inference speed. |
| `agnostic_nms` | `bool` | `False` | Enables class-agnostic Non-Maximum Suppression(NMS),whichmergesoverlappingboxesofdifferent classes. Useful in multi-class detection scenarios where class overlap is common. |
| `classes` | `list[int]` | `None` | Filters predictions to a set of class IDs. Only detections belonging to the specified classes will be returned. Useful for focusing on relevant objects in multi-class detection tasks. |
| `retina_masks` | `bool` | `False` | Uses high-resolution segmentation masks if available in the model. This can enhance mask quality for segmentation tasks, providing finer detail. |
| `embed` | `list[int]` | `None` | Specifies the layers from which to extract feature vectors or embeddings. Useful for downstream tasks like clustering or similarity search. |
| `show` | `bool` | `False` | If `True`, displays the annotated images or videos in a window. Useful for immediate visual feedback during development or testing. |
| `save` | `bool` | `False` | Enables saving of the annotated images or videos to file. Useful for documentation, further analysis, or sharing results. |
| `save_frames` | `bool` | `False` | When processing videos, saves individual frames as images. Useful for extracting specific frames or for detailed frame-by-frame analysis. |
| `save_txt` | `bool` | `False` | Saves detection results in a text file, following the format `[class] [x_center] [y_center] [width] [height] [confidence]`. Useful for integration with other analysis tools. |
| `save_conf` | `bool` | `False` | Includes confidence scores in the saved text files. Enhances the detail available for post-processing and analysis. |
| `save_crop` | `bool` | `False` | Saves cropped images of detections. Useful for dataset augmentation, analysis, or creating focused datasets for specific objects. |
| `show_labels` | `bool` | `True` | Displays labels for each detection in the visual output. Provides immediate understanding of detected objects. |
| `show_conf` | `bool` | `True` | Displays the confidence score for each detection alongside the label. Gives insight into the model's certainty for each detection. |
| `show_boxes` | `bool` | `True` | Draws bounding boxes around detected objects. Essential for visual identification and location of objects in images or video frames. |
| `line_width` | `None or int` | `None` | Specifies the line width of bounding boxes. If `None`, the line width is automatically adjusted based on the image size. Provides visual customization for clarity. |
Glenn Jocher
12 months ago
committed by
GitHub