Merge pull request #7 from 5sControl/dev

Dev

Author: eugenos-programos

.gitignore

@@ -6,4 +6,5 @@ __pycache__
 .env
 .vscode/
 build/
-cmake-build-debug/
+cmake-build-debug/
+idle_models/weights/

confs/configs.json

@@ -1,6 +1,6 @@
 {
     "classes": [
-        67
+        77
     ],
     "wait_time": 10,
     "threshold": 100

confs/settings.env

@@ -1,5 +1,5 @@
 username = 'admin'
 password = 'just4Taqtile'
-camera_url = 'http://192.168.1.162/onvif-http/snapshot?Profile_1'
+camera_url = 'http://192.168.1.110:3456/onvif-http/snapshot?camera_ip=192.168.1.168'
 server_url = 'http://127.0.0.1'
 folder = 'images/192.168.1.163'

idle_models/Dockerfile

@@ -9,4 +9,4 @@ WORKDIR /var/www/5scontrol
 COPY . .
 RUN mkdir -p /usr/src/app
 EXPOSE 5001
-CMD [ "flask", "run","--host","0.0.0.0","--port","5001"]
+CMD [ "flask", "run","--host","0.0.0.0","--port", "5001"]

idle_models/IdleObjectDetectionModel.py

@@ -1,34 +1,26 @@
 import torch
-from yolor.model import get_model
 import numpy as np
-from yolor.utils.datasets import letterbox
-from yolor.utils.general import non_max_suppression, scale_coords
+from transformers import DetrImageProcessor, DetrForObjectDetection, DetrConfig
+import torch
 
 
 class IdleObjectDetectionModel:
     def __init__(self, path: str, conf_thresh, iou_thresh, classes) -> None:
-        self.model, self.device = get_model(path, "yolor/yolor_csp_x.cfg")
+        self.model = DetrForObjectDetection.from_pretrained(path)
+        self.model.eval()
+        self.processor = DetrImageProcessor.from_pretrained(path)
         self.conf_thresh = conf_thresh
         self.iou_thresh = iou_thresh
         self.classes = classes
 
-    def __preprocess_image__(self, img: np.array) -> np.array:
-        self.img_shape = img.shape
-        img = letterbox(img.copy(), new_shape=1280, auto_size=64)[0]
-        img = img[:, :, ::-1].transpose(2, 0, 1)
-        img = np.ascontiguousarray(img)
-        img = torch.from_numpy(img).to(self.device)
-        img = img.float()
-        img /= 255.0
-        img = img.unsqueeze(0)
-        return img
-
     @torch.no_grad()
-    def __call__(self, img: np.array) -> torch.Tensor:
-        img = self.__preprocess_image__(img)
-        pred = self.model(img, augment=False)[0]
-        pred = non_max_suppression(
-            pred, 0.45, 0.5, classes=[67], agnostic=False)[0]
-        pred[:, :4] = scale_coords(
-            img.shape[2:], pred[:, :4], self.img_shape).round()
-        return pred[:, :5]
+    def __call__(self, img: np.array) -> np.array:
+        inputs = self.processor(images=img, return_tensors="pt")
+        outputs = self.model(**inputs)
+        target_sizes = torch.tensor([img.shape[:-1]])
+        results = self.processor.post_process_object_detection(outputs, target_sizes=target_sizes, threshold=self.conf_thresh)[0]
+        pred = []
+        for score, label, box in zip(results["scores"], results["labels"], results["boxes"]):
+            if self.model.config.id2label[label.item()] == 'cell phone':
+                pred.append([round(i, 2) for i in box.tolist()] + [score.item()])
+        return np.array(pred)

idle_models/app.py

@@ -4,6 +4,8 @@
 from IdleObjectDetectionModel import IdleObjectDetectionModel
 import numpy as np
 import io
+import colorlog
+import logging
 
 
 app = Flask(__name__)
@@ -14,13 +16,29 @@
     CLASSES
 )
 
+logger = logging.getLogger('min_max_logger')
+handler = colorlog.StreamHandler()
+handler.setFormatter(colorlog.ColoredFormatter(
+        '%(log_color)s%(asctime)s %(levelname)s: %(message)s',
+        datefmt='%Y-%m-%d %H:%M:%S',
+        log_colors={
+            'DEBUG': 'cyan',
+            'INFO': 'green',
+            'WARNING': 'yellow',
+            'CRITICAL': 'bold_red,bg_white',
+        }))
+logger.addHandler(handler)
+logger.setLevel(logging.DEBUG)
+logger.propagate = False
+
 convert_bytes2image = lambda bytes: np.array(Image.open(io.BytesIO(bytes)), dtype=np.uint8)
 
 @app.route('/predict', methods=['POST'])
 def predict():
     if request.method == 'POST':
         image = convert_bytes2image(request.files["image"].read())
         coords = model(image)
+        logger.info(f"Request to predict - {len(coords)}")
         return jsonify(
             {
                 "coordinates": coords.tolist()

idle_models/configs/confs.json

@@ -3,7 +3,7 @@
         67
     ],
     "iou_thres": 0.5,
-    "conf_thres": 0.5,
-    "model_path": "yolor/yolor_csp_x_star.pt",
+    "conf_thres": 0.65,
+    "model_path": "./weights",
     "port": 5001
 }

idle_models/requirements.txt

@@ -7,4 +7,7 @@ pydantic==1.10.2
 python-dotenv==1.0.0
 PyYAML==6.0
 requests==2.27.1
-Flask==2.2.2
+Flask==2.2.2
+colorlog==4.8.0
+transformers==4.28.1
+timm==0.6.13