Merge pull request #15 from 5sControl/dev

Dev

Author: eugenos-programos

.dockerignore

@@ -1,3 +1,4 @@
 .cache
 __pycache__
-idle_models
+idle_models
+.git

.gitignore

@@ -3,7 +3,7 @@
 images/
 env/
 __pycache__
-.env
+*.env
 .vscode/
 build/
 cmake-build-debug/

IdleAlgorithm.py

@@ -21,28 +21,31 @@ def __init__(
         self._reporter = reporter
         self._min_epoch_time = 2 # replace on config in future
 
-    async def start(self) -> None:
+    def start(self) -> None:
+        self.prev_preds = np.array([[]]).astype(np.float32)
         while True:
             start_epoch_time = time.time()
             self._run_one_idle_epoch()
             end_epoch_time = time.time()
             passed_time = end_epoch_time - start_epoch_time
             if passed_time < self._min_epoch_time:
-                await asyncio.sleep(self._min_epoch_time - passed_time)
+                time.sleep(self._min_epoch_time - passed_time)
 
     def _run_one_idle_epoch(self) -> None:
             img, start_tracking = self._image_extractor.get_snapshot()
             if img is None:
                 return
+            self._logger.debug("Sending request for model predictions")
             preds = self._model_predictor.predict(img)
+            self._logger.debug("Model predictiions was recieved")
             if preds is None:
                 return
             if preds.size != 0 and not np.any(preds == 1.):
                 self._logger.info("Telephone is detected")
-                if utils.bboxes_not_equal(prev_preds, preds, configs["threshold"]):
+                if utils.bboxes_not_equal(self.prev_preds, preds, configs["threshold"]):
                     utils.save_cropped_bbox(img, preds[:, :4])
                     img = utils.put_rectangle(img, preds[:, :4], preds[:, 4])
                     self._reporter.send_report(self._reporter.create_report(img, str(start_tracking)))
                 else:
                     self._logger.debug("Equal bboxes")
-                prev_preds = preds
+                self.prev_preds = preds

connection/ImageHTTPExtractor.py

@@ -4,11 +4,7 @@
 import numpy as np
 import cv2
 from typing import Tuple, Union
-import socketio
-
-
-sio = socketio.AsyncClient()
-images = {}
+import requests
 
 
 class ImageHTTPExtractor:
@@ -20,27 +16,12 @@ def __init__(self, camera_ip: str, logger: logging.Logger, **credentials) -> Non
 
     def get_snapshot(self) -> Tuple[Union[cv2.Mat, None], Union[datetime.time, None]]:
         try:
-            global images
-            image = images[self.camera_ip]            
+            response = requests.get(self.camera_ip)  
             curr_time = datetime.datetime.now()
-            nparr = np.frombuffer(image, np.uint8)
+            nparr = np.frombuffer(response.content, np.uint8)
             img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
             return img, curr_time
         except Exception as exc:
             self.logger.warning(f"Cannot retrieve image. Following error raised - {exc}")
-            print(images.keys())
             return None, None
 
-@sio.event
-async def connect():
-    print("Connection")
-
-@sio.event
-async def snapshot_updated(data):
-    camera_ip, screen = data.get("camera_ip"), data.get("screenshot")
-    global images
-    images[camera_ip] = screen
-
-async def run_sio(url):
-    await sio.connect(url)
-    await sio.wait()

connection/__init__.py

@@ -1,3 +1,3 @@
 from .IdleReporter import IdleReporter
-from .ImageHTTPExtractor import ImageHTTPExtractor, run_sio
+from .ImageHTTPExtractor import ImageHTTPExtractor
 from .ModelPredictionsReceiver import ModelPredictionsReceiver

idle_model/IdleObjectDetectionModel.py

@@ -3,14 +3,16 @@
 import numpy as np
 from yolor.utils.datasets import letterbox
 from yolor.utils.general import non_max_suppression, scale_coords
+import time
 
 
 class IdleObjectDetectionModel:
-    def __init__(self, model_path: str, config_path: str, conf_thresh: float, iou_thresh: float, classes: list) -> None:
+    def __init__(self, model_path: str, config_path: str, conf_thresh: float, iou_thresh: float, classes: list, logger) -> None:
         self.model, self.device = get_model(model_path, config_path)
         self.conf_thresh = conf_thresh
         self.iou_thresh = iou_thresh
         self.classes = classes
+        self.logger = logger
 
     def __preprocess_image__(self, img: np.array) -> np.array:
         self.img_shape = img.shape
@@ -26,7 +28,9 @@ def __preprocess_image__(self, img: np.array) -> np.array:
     @torch.no_grad()
     def __call__(self, img: np.array) -> list:
         img = self.__preprocess_image__(img)
+        t1 = time.time()
         pred = self.model(img, augment=False)[0]
+        self.logger.info(f"Time used for predictions - {time.time() - t1}")
         pred = non_max_suppression(pred, self.conf_thresh, self.iou_thresh, classes=[67], agnostic=False)[0]
         pred[:, :4] = scale_coords(img.shape[2:], pred[:, :4], self.img_shape).round()
         return pred[:, :5]

idle_model/app.py

@@ -8,15 +8,6 @@
 import logging
 
 
-app = Flask(__name__)
-model = IdleObjectDetectionModel(
-    MODEL_PATH,
-    CONF_PATH,
-    CONF_THRES,
-    IOU_THRES,
-    CLASSES
-)
-
 logger = logging.getLogger('min_max_logger')
 handler = colorlog.StreamHandler()
 handler.setFormatter(colorlog.ColoredFormatter(
@@ -32,6 +23,16 @@
 logger.setLevel(logging.DEBUG)
 logger.propagate = False
 
+app = Flask(__name__)
+model = IdleObjectDetectionModel(
+    MODEL_PATH,
+    CONF_PATH,
+    CONF_THRES,
+    IOU_THRES,
+    CLASSES,
+    logger
+)
+
 convert_bytes2image = lambda bytes: np.array(Image.open(io.BytesIO(bytes)), dtype=np.uint8)
 
 @app.route('/predict', methods=['POST'])

main.py

@@ -1,5 +1,5 @@
 import utils
-from connection import run_sio, IdleReporter, ImageHTTPExtractor, ModelPredictionsReceiver
+from connection import IdleReporter, ImageHTTPExtractor, ModelPredictionsReceiver
 from confs import configs
 from dotenv import load_dotenv
 from IdleAlgorithm import IdleAlgorithm
@@ -14,7 +14,7 @@
 password = os.environ.get("password")
 username = os.environ.get("username")
 server_url = os.environ.get("server_url")
-camera_ip = os.environ.get("camera_ip")
+camera_ip = os.environ.get("camera_url")
 folder = os.environ.get("folder")
 
 logger = utils.create_logger()
@@ -25,13 +25,5 @@
 model_predictor = ModelPredictionsReceiver(server_url, logger)
 algo = IdleAlgorithm(logger, image_extractor, model_predictor, reporter)
 
-async def main():
-    await asyncio.gather(run_sio(server_url + ':3456'), algo.start())
 
-loop = asyncio.get_event_loop()
-try:
-    loop.run_until_complete(main())
-except Exception as exc:
-    print(exc)
-finally:
-    loop.close()
+algo.start()

requirements.txt

@@ -1,13 +1,9 @@
-colorlog==4.8.0
+colorlog==6.7.0
 httplib2==0.22.0
 numpy==1.22.3
 opencv_python==4.7.0.72
 pafy==0.5.5
-Pillow==9.5.0
 pydantic==1.10.2
 python-dotenv==1.0.0
-PyYAML==6.0
 requests==2.27.1
-numba==0.57.1
-python-socketio==5.8.0
-aiohttp==3.8.1
+numba==0.57.1