Spaces:
Running
Running
File size: 1,894 Bytes
b87a24a c1bf807 b87a24a c1bf807 b87a24a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 | import cv2
import multiprocessing as mp
BASE_IMG_SIZE = 640
REF_PIXELS = 640 * 640
REF_FPS_CPU = 13.0
TRACK_STABILITY_STRIDE = 3
def _cpu_score():
return mp.cpu_count()
def _video_meta(path):
cap = cv2.VideoCapture(path)
fps = cap.get(cv2.CAP_PROP_FPS)
frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
cap.release()
duration = frames / fps if fps else 0
pixels = w * h
return fps, frames, duration, w, h, pixels
def _estimate_fps(imgsz, cpu_score):
scale = (imgsz * imgsz) / REF_PIXELS
return (REF_FPS_CPU * cpu_score / 12) / scale
def _select_imgsz(pixels):
if pixels >= 3840 * 2160:
return 640
if pixels >= 2560 * 1440:
return 704
if pixels >= 1920 * 1080:
return 736
if pixels >= 1280 * 720:
return 800
return 960
def _select_stride(video_fps, model_fps):
if model_fps >= video_fps:
return 1
ratio = video_fps / model_fps
stride = int(round(ratio))
return max(1, min(stride, TRACK_STABILITY_STRIDE))
def get_optimal_config(video_path):
fps, frames, duration, w, h, pixels = _video_meta(video_path)
cpu_score = _cpu_score()
imgsz = _select_imgsz(pixels)
model_fps = _estimate_fps(imgsz, cpu_score)
detect_stride = _select_stride(fps, model_fps)
effective_fps = model_fps / detect_stride
realtime_possible = effective_fps >= fps
return {
"video_fps": fps,
"frames": frames,
"duration": round(duration, 2),
"resolution": [w, h],
"pixels": pixels,
"cpu_score": cpu_score,
"imgsz": imgsz,
"detect_stride": detect_stride,
"model_fps_est": round(model_fps, 2),
"effective_fps_est": round(effective_fps, 2),
"realtime_possible": realtime_possible
}
|