Add object detection, tracking, mobile notifications, and search to any security camera.
__main__ so beam works
Rory Cleardf3240743cc7
diff --git a/experimental/rtsp_yolov8.py b/experimental/rtsp_yolov8.pyindex abb3a43..d2cf30b 100644--- a/experimental/rtsp_yolov8.py+++ b/experimental/rtsp_yolov8.py@@ -12,6 +12,11 @@ import time, sysfrom tinygrad.helpers import fetchfrom tinygrad.nn.state import safe_load, load_state_dictimport json+import cv2+import time+from http.server import BaseHTTPRequestHandler, HTTPServer+from socketserver import ThreadingMixIn+import threading#Model architecture from https://github.com/ultralytics/ultralytics/issues/189#The upsampling class has been taken from this pull request https://github.com/tinygrad/tinygrad/pull/784 by dc-dc-dc. Now 2(?) models use upsampling. (retinet and this)@@ -323,22 +328,6 @@ def do_inf(im):predictions = yolo_infer(im)return predictions-# Model initialization (same as before)-yolo_variant = sys.argv[2] if len(sys.argv) >= 3 else (print("No variant given, so choosing 'n' as the default. Yolov8 has different variants, you can choose from ['n', 's', 'm', 'l', 'x']") or 'n')-depth, width, ratio = get_variant_multiples(yolo_variant)-yolo_infer = YOLOv8(w=width, r=ratio, d=depth, num_classes=80)-state_dict = safe_load(get_weights_location(yolo_variant))-load_state_dict(yolo_infer, state_dict)--import cv2-import time-from http.server import BaseHTTPRequestHandler, HTTPServer-from socketserver import ThreadingMixIn-import threading--class_labels = fetch('https://raw.githubusercontent.com/pjreddie/darknet/master/data/coco.names').read_text().split("\n")-color_dict = {label: tuple((((i+1) * 50) % 256, ((i+1) * 100) % 256, ((i+1) * 150) % 256)) for i, label in enumerate(class_labels)}-# RTSP URL# Video capture threadclass VideoCapture:@@ -432,13 +421,23 @@ class ThreadedHTTPServer(ThreadingMixIn, HTTPServer):# --- Start server ---rtsp_url = sys.argv[1] if len(sys.argv) >= 2 else (print("No rtsp url given") or sys.exit(1))-cam = VideoCapture(rtsp_url)-try:+if __name__ == "__main__":+ # Model initialization (same as before)+ yolo_variant = sys.argv[2] if len(sys.argv) >= 3 else (print("No variant given, so choosing 'n' as the default. Yolov8 has different variants, you can choose from ['n', 's', 'm', 'l', 'x']") or 'n')+ depth, width, ratio = get_variant_multiples(yolo_variant)+ yolo_infer = YOLOv8(w=width, r=ratio, d=depth, num_classes=80)+ state_dict = safe_load(get_weights_location(yolo_variant))+ load_state_dict(yolo_infer, state_dict)+ class_labels = fetch('https://raw.githubusercontent.com/pjreddie/darknet/master/data/coco.names').read_text().split("\n")+ color_dict = {label: tuple((((i+1) * 50) % 256, ((i+1) * 100) % 256, ((i+1) * 150) % 256)) for i, label in enumerate(class_labels)}+ cam = VideoCapture(rtsp_url)++ try:server = ThreadedHTTPServer(('0.0.0.0', 8080), MJPEGHandler)print("\nServing video on http://<this-ip>:8080")server.serve_forever()-except KeyboardInterrupt:+ except KeyboardInterrupt:print("\nShutting down...")cam.release()server.shutdown()