Merge pull request #34 from 5sControl/dev

Dev

connection/ImageCapture.py

@@ -11,7 +11,6 @@ def __init__(self, path, **kwargs):
         self._username = kwargs.get('username')
         self._password = kwargs.get('password')
         self._logger : Logger = kwargs.get('logger')
-        self._server_url = kwargs.get("server_url")
         self._prev_img = None
         if not self._username or not self._password:
             self.logger.warning("Empty password or username")
@@ -29,5 +28,6 @@ def get_snapshot(self) -> cv2.Mat:
             self._prev_img = image
             return image, ssim_value
         except Exception as exc:
+            print(self._camera_ip)
             self._logger.warning(f"Empty image.\n {exc} \n Skipping iteration...")
             return None, None

main.py

@@ -19,13 +19,13 @@
 username = os.environ.get("username")
 password = os.environ.get("password")
 server_url = os.environ.get("server_url")
-camera_ip = os.environ.get("camera_url")
+camera_url = os.environ.get("camera_url")
 folder = os.environ.get("folder")
 
 logger = create_logger()
 
 dataset = ImageCapture(
-    camera_ip,
+    camera_url,
     username=username, 
     password=password, 
     logger=logger,

min_max_utils/MinMaxReporter.py

@@ -173,6 +173,6 @@ def send_report_to_server(self, report: dict) -> None:
                        f'{self.server_url}:8000/api/reports/report-with-photos/'])
         )
         try:
-            requests.post(url=f'{self.server_url}:80/api/reports/report-with-photos/', json=report)
+            requests.post(url=os.environ.get("link_reports"), json=report)
         except Exception as exc:
             self.logger.error("Error while sending report occurred: {}".format(exc))

min_max_utils/min_max_utils.py

@@ -103,15 +103,12 @@ def most_common(lst):
 
 
 def check_box_in_area(box_coord: list, area_coord: list):
-    box_corners = (
-        (box_coord[0], box_coord[1]),
-        (box_coord[0], box_coord[3]),
-        (box_coord[2], box_coord[1]),
-        (box_coord[2], box_coord[3])
-    )
-    for box_corner in box_corners:
-        if area_coord[0] < box_corner[0] <= area_coord[2] and area_coord[1] < box_corner[1] <= area_coord[3]:
-            return True
+    box_center = [
+        (box_coord[0] + box_coord[2]) / 2, 
+        (box_coord[1] + box_coord[3]) / 2
+    ]
+    if area_coord[0] < box_center[0] < area_coord[2] and area_coord[1] < box_center[1] < area_coord[3]:
+        return True
     return False
 
 

model_image/Dockerfile

@@ -1,4 +1,4 @@
-FROM python:3.10
+FROM python:3.9
 RUN apt-get update
 RUN pip install torch==1.13.1+cpu torchvision==0.14.1+cpu --extra-index-url https://download.pytorch.org/whl/cpu
 COPY requirements.txt .
@@ -9,4 +9,4 @@ WORKDIR /var/www/5scontrol
 COPY . .
 RUN mkdir -p /usr/src/app
 EXPOSE 5000
-CMD [ "flask", "run","--host","0.0.0.0","--port","5000"]
+CMD ["flask", "run", "--host", "0.0.0.0", "--port", "5000"]

model_image/YOLONASObjDetectionModel.py

@@ -0,0 +1,26 @@
+from super_gradients.training import models
+from super_gradients.common.object_names import Models
+import torch
+import numpy as np
+
+
+class YOLONASObjDetectionModel:
+    def __init__(self, model_path: str, conf_thresh: float, iou_thresh: float, classes: list, img_size: int) -> None:
+        # self.device = torch.device("cuda:0") if torch.cuda.is_available() else torch.device("cpu")
+        self._model = models.get(model_path, pretrained_weights='coco', num_classes=80)
+        self._model.eval()
+        self.conf_thresh = conf_thresh
+        self.iou_thresh = iou_thresh
+        self.classes = classes
+        self.img_size = img_size
+
+    @torch.no_grad()
+    def __call__(self, img: np.array) -> list:
+        results = self._model.predict(img, fuse_model=False)
+        prediction = results[0].prediction
+        labels = prediction.labels.astype(int) == self.classes[0]
+        boxes = prediction.bboxes_xyxy[labels]
+        confs = np.expand_dims(prediction.confidence.astype(float)[labels], -1)
+        #confs = 
+        coords_with_confs = np.hstack((boxes, confs))
+        return coords_with_confs

model_image/app.py

@@ -2,6 +2,7 @@
 from flask import Flask, jsonify, request
 from flask_configs.load_configs import bottle_model_configs, box_model_configs, human_model_configs
 from ObjectDetectionModel import YOLOv8ObjDetectionModel
+from YOLONASObjDetectionModel import YOLONASObjDetectionModel
 import numpy as np
 import colorlog
 import logging
@@ -23,9 +24,9 @@
 logger.propagate = False
 
 app = Flask(__name__)
-human_model = YOLOv8ObjDetectionModel(**human_model_configs)
+human_model = YOLONASObjDetectionModel(**human_model_configs)
 box_model = YOLOv8ObjDetectionModel(**box_model_configs)
-bottle_model = YOLOv8ObjDetectionModel(**bottle_model_configs)
+bottle_model = YOLONASObjDetectionModel(**bottle_model_configs)
 
 
 convert_bytes2image = lambda bytes: np.array(Image.open(io.BytesIO(bytes)), dtype=np.uint8)

model_image/flask_configs/flask_confs.json

@@ -14,7 +14,7 @@
         ],
         "iou_thresh": 0.5,
         "conf_thresh": 0.5,
-        "model_path": "weigths/min_max_v0.5.2_human.pt",
+        "model_path": "yolo_nas_s",
         "img_size": 640
     },
     "bottle_model": {
@@ -23,7 +23,7 @@
         ],
         "iou_thresh": 0.5,
         "conf_thresh": 0.5,
-        "model_path": "weigths/min_max_v0.5.2_bottle.pt",
+        "model_path": "yolo_nas_m",
         "img_size": 640
     },
     "port": 5000

model_image/requirements.txt

@@ -1,8 +1,7 @@
-colorlog==4.8.0
-httplib2==0.22.0
-numpy==1.22.3
-opencv_python==4.7.0.72
-ultralytics==8.0.136
+ultralytics==8.0.191
+super-gradients==3.2.1
+timm==0.6.7
 Flask==2.2.2
-py-cpuinfo==9.0.0
-onnxruntime==1.15.1
+numpy==1.23.0
+Werkzeug==2.2.2
+colorlog==4.8.0