1. 矫正矩阵

2. 添加一个启动脚本

src/grpc/base_cfg.py

@@ -0,0 +1,63 @@
+# -*- coding: utf-8 -*-
+'''
+@Author: captainfffsama
+@Date: 2023-02-02 16:40:37
+@LastEditors: captainfffsama tuanzhangsama@outlook.com
+@LastEditTime: 2023-02-02 16:42:41
+@FilePath: /LoFTR/src/grpc/base_cfg.py
+@Description:
+'''
+import json
+import yaml
+
+param = dict(grpc=dict(host='127.0.0.1',
+                       port='8001',
+                       max_workers=10,
+                       max_send_message_length=100 * 1024 * 1024,
+                       max_receive_message_length=100 * 1024 * 1024),
+             loftr=dict(ckpt_path="",device="cuda:0",thr=0.5,ransc_method="USAC_MAGSAC",ransc_thr=3,ransc_max_iter=2000,
+))
+
+
+def _update(dic1: dict, dic2: dict):
+    """使用dic2 来递归更新 dic1
+        # NOTE:
+        1. dic1 本体是会被更改的!!!
+        2. python 本身没有做尾递归优化的,dict深度超大时候可能爆栈
+    """
+    for k, v in dic2.items():
+        if k.endswith('args') and v is None:
+            dic2[k] = {}
+        if k in dic1:
+            if isinstance(v, dict) and isinstance(dic1[k], dict):
+                _update(dic1[k], dic2[k])
+            else:
+                dic1[k] = dic2[k]
+        else:
+            dic1[k] = dic2[k]
+
+
+def _merge_yaml(yaml_path: str):
+    global param
+    with open(yaml_path, 'r') as fr:
+        content_dict = yaml.load(fr, yaml.FullLoader)
+    _update(param, content_dict)
+
+
+def _merge_json(json_path: str):
+    global param
+    with open(json_path, 'r') as fr:
+        content_dict = json.load(fr)
+    _update(param, content_dict)
+
+
+def merge_param(file_path: str):
+    """按照用户传入的配置文件更新基本设置
+    """
+    cfg_ext = file_path.split('.')[-1]
+    func_name = '_merge_' + cfg_ext
+    if func_name not in globals():
+        raise ValueError('{} is not support'.format(cfg_ext))
+    else:
+        globals()[func_name](file_path)
+

src/grpc/loftr_worker.py

@@ -12,9 +12,10 @@ import numpy as np
 
 from src.loftr import LoFTR, default_cfg
 
-DebugInfo=namedtuple("DebugInfo",
-                        ["kp0_fake_match","kp1_fake_match",
-                        "kp0_true_match","kp1_true_match"])
+DebugInfo = namedtuple(
+    "DebugInfo",
+    ["kp0_fake_match", "kp1_fake_match", "kp0_true_match", "kp1_true_match"])
+
 
 class LoFTRWorker(object):
 
@@ -32,20 +33,27 @@ class LoFTRWorker(object):
             device = 'cpu'
             print("ERROR: cuda can not use, will use cpu")
         self.model = self.model.eval().to(device)
-        self.thr=thr
-        self.ransc_method = getattr(cv2,ransc_method)
-        self.ransc_thr=ransc_thr
-        self.ransc_max_iter=ransc_max_iter
+        self.thr = thr
+        self.ransc_method = getattr(cv2, ransc_method)
+        self.ransc_thr = ransc_thr
+        self.ransc_max_iter = ransc_max_iter
 
-    def _img2gray(self, img):
+    def _imgdeal(self, img):
         if len(img.shape) == 3 and img.shape[-1] == 3:
             img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
 
-        return img
+        oh, ow = img.shape[:2]
+        img = cv2.resize(img, (640, 480))
+        h, w = img.shape[:2]
+        fix_matrix = np.array([[w / ow, 0, 0], [0, h / oh, 0], [0, 0, 1]])
+        return img, fix_matrix
+
+    def _fix_H(self, fm0, fm1, H):
+        return np.linalg.inv(fm0) @ H @ fm1
 
-    def __call__(self, img0, img1,debug=""):
-        img0 = self._img1gray(img0)
-        img1 = self._img1gray(img1)
+    def __call__(self, img0, img1, debug=""):
+        img0, fm0 = self._imgdeal(img0)
+        img1, fm1 = self._imgdeal(img1)
         img0 = torch.from_numpy(img0)[None][None].cuda() / 255.
         img1 = torch.from_numpy(img1)[None][None].cuda() / 255.
 
@@ -56,33 +64,33 @@ class LoFTRWorker(object):
             mkpts1 = batch['mkpts1_f'].cpu().numpy()
             mconf = batch['mconf'].cpu().numpy()
 
-        idx=np.where(mconf>self.thr)
-        mconf=mconf[idx]
-        mkpts0=mkpts0[idx]
-        mkpts1=mkpts1[idx]
+        idx = np.where(mconf > self.thr)
+        mconf = mconf[idx]
+        mkpts0 = mkpts0[idx]
+        mkpts1 = mkpts1[idx]
 
-        debug_info=None
+        debug_info = None
+        H = np.array([[1, 0, 0], [0, 1, 0], [0, 0, 1]], dtype=np.float)
         if mkpts0.shape[0] < 4 or mkpts1.shape[0] < 4:
-            return np.array([[1, 0, 0], [0, 1, 0], [0, 0, 1]],
-                            dtype=np.float), False,debug_info
+            return self._fix_H(fm0, fm1, H), False, debug_info
 
         H, Mask = cv2.findHomography(mkpts0[:, :2],
-                                    mkpts1[:, :2],
-                                    self.ransc_method,
-                                    self.ransc_thr,
-                                    maxIters=self.ransc_max_iter)
-        Mask=np.squeeze(Mask)
+                                     mkpts1[:, :2],
+                                     self.ransc_method,
+                                     self.ransc_thr,
+                                     maxIters=self.ransc_max_iter)
+        Mask = np.squeeze(Mask)
         if debug:
 
-            kp0_true_matched=mkpts0[Mask.astype(bool),:2]
-            kp1_true_matched=mkpts1[Mask.astype(bool),:2]
-            kp0_fake_matched=mkpts0[~Mask.astype(bool),:2]
-            kp1_fake_matched=mkpts1[~Mask.astype(bool),:2]
+            kp0_true_matched = mkpts0[Mask.astype(bool), :2]
+            kp1_true_matched = mkpts1[Mask.astype(bool), :2]
+            kp0_fake_matched = mkpts0[~Mask.astype(bool), :2]
+            kp1_fake_matched = mkpts1[~Mask.astype(bool), :2]
 
-            debug_info=DebugInfo(kp0_fake_matched,kp1_fake_matched,kp0_true_matched,kp1_true_matched)
+            debug_info = DebugInfo(kp0_fake_matched, kp1_fake_matched,
+                                   kp0_true_matched, kp1_true_matched)
 
         if H is None:
-            return np.array([[1, 0, 0], [0, 1, 0], [0, 0, 1]],
-                            dtype=np.float), False,debug_info
+            return self._fix_H(fm0, fm1, H), False, debug_info
         else:
-            return H, True,debug_info
+            return self._fix_H(fm0, fm1, H), True, debug_info

src/grpc/server.py

@@ -2,12 +2,14 @@
 @Author: captainfffsama
 @Date: 2023-02-02 15:59:46
 @LastEditors: captainfffsama tuanzhangsama@outlook.com
-@LastEditTime: 2023-02-02 16:08:41
+@LastEditTime: 2023-02-02 16:43:55
 @FilePath: /LoFTR/src/grpc/server.py
 @Description:
 '''
 import numpy as np
 
+from src.loftr import default_cfg
+
 from . import loftr_pb2
 from .loftr_pb2_grpc import LoftrServicer
 
@@ -17,7 +19,6 @@ from .utils import decode_img_from_proto, np2tensor_proto, img2pb_img
 
 class LoFTRServer(LoftrServicer):
     def __init__(self,
-                 config,
                  ckpt_path,
                  device="cuda:0",
                  thr=0.5,
@@ -28,7 +29,7 @@ class LoFTRServer(LoftrServicer):
                  **kwargs):
         super().__init__(*args, **kwargs)
         self.worker = LoFTRWorker(
-            config,
+            default_cfg,
             ckpt_path,
             device,
             thr,

start_grpc.py

@@ -0,0 +1,62 @@
+# -*- coding: utf-8 -*-
+'''
+@Author: captainfffsama
+@Date: 2023-02-02 16:38:45
+@LastEditors: captainfffsama tuanzhangsama@outlook.com
+@LastEditTime: 2023-02-02 16:48:12
+@FilePath: /LoFTR/start_grpc.py
+@Description:
+'''
+from concurrent import futures
+import sys
+from pprint import pprint
+import os
+import yaml
+
+import grpc
+
+from src.grpc.loftr_pb2_grpc import add_LoftrServicer_to_server
+import src.grpc.base_cfg as cfg
+from src.grpc.server import LoFTRServer
+
+
+def start_server(config):
+    if not os.path.exists(config):
+        raise FileExistsError('{} 不存在'.format(config))
+    cfg.merge_param(config)
+    args_dict: dict = cfg.param
+    pprint(args_dict)
+
+    grpc_args = args_dict['grpc']
+    model_args = args_dict['loftr']
+    #  最大限制为100M
+    server = grpc.server(
+        futures.ThreadPoolExecutor(max_workers=grpc_args['max_workers']),
+        options=[('grpc.max_send_message_length',
+                  grpc_args['max_send_message_length']),
+                 ('grpc.max_receive_message_length',
+                  grpc_args['max_receive_message_length'])])
+
+    loftr_server =LoFTRServer(**model_args)
+    add_LoftrServicer_to_server(loftr_server,server)
+    server.add_insecure_port("{}:{}".format(grpc_args['host'],
+                                            grpc_args['port']))
+    server.start()
+    server.wait_for_termination()
+
+
+def main(args=None):
+    import argparse
+    parser = argparse.ArgumentParser(description="grpc调用loftr,需要配置文件")
+    parser.add_argument("-c", "--config", type=str, default="", help="配置文件地址")
+    options = parser.parse_args(args)
+    if options.config:
+        start_server(options.config)
+
+if __name__ == "__main__":
+    rc = 1
+    try:
+        main()
+    except Exception as e:
+        print('Error: %s' % e, file=sys.stderr)
+    sys.exit(rc)