app.py

import cv2
import gradio as gr
import supervision as sv
from ultralytics import YOLO
from PIL import Image
import torch
import time
import numpy as np
import uuid
import spaces

ver=[0,0,0,0,0,0,6,7,8,9,10,11]
ltr=["n","s","m","1","x"]
tsk=["","-seg","-pose","-obb","-cls"]
#yolov8s.pt
modin=f"yolov{ver[9]}{ltr[1]}{tsk[0]}.pt"
model = YOLO(modin)
annotators = ["Box","RoundBox","BoxCorner","Color",
              "Circle","Dot","Triangle","Elipse","Halo",
              "PercentageBar","Mask","Polygon","Label",
              "RichLabel","Icon","Crop","Blur","Pixelate","HeatMap"]

@spaces.GPU
def stream_object_detection(video):
    SUBSAMPLE=1
    cap = cv2.VideoCapture(video)
    # This means we will output mp4 videos
    video_codec = cv2.VideoWriter_fourcc(*"mp4v") # type: ignore
    fps = int(cap.get(cv2.CAP_PROP_FPS))
    desired_fps = fps // SUBSAMPLE
    width  = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)) // 2
    height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)) // 2
    iterating, frame = cap.read()
    n_frames = 0
    output_video_name = f"output_{uuid.uuid4()}.mp4"
    output_video = cv2.VideoWriter(output_video_name, video_codec, desired_fps, (width, height)) # type: ignore

    while iterating:
        frame = cv2.resize( frame, (0,0), fx=0.5, fy=0.5)
        frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
        result = model(Image.fromarray(frame))[0]
        detections = sv.Detections.from_ultralytics(result)
        print(detections)

        box_annotator = eval(f'sv.{annotators[0]}Annotator()')
        
        outp = box_annotator.annotate(
          scene=frame.copy(),
          detections=detections)
        
        #outp = draw_box(frame,detections)
        frame = np.array(outp)
        # Convert RGB to BGR
        frame = frame[:, :, ::-1].copy()
        output_video.write(frame)
        batch = []
        output_video.release()
        yield output_video_name,detections
        output_video_name = f"output_{uuid.uuid4()}.mp4"
        output_video = cv2.VideoWriter(output_video_name, video_codec, desired_fps, (width, height)) # type: ignore
        iterating, frame = cap.read()
        n_frames += 1

with gr.Blocks() as app:
    gr.HTML("<div style='font-size: 50px;font-weight: 800;'>Supervision</div><div style='font-size: 30px;'>Video Object Detection</div><div>Github:<a href='https://github.com/roboflow/supervision' target='_blank'>https://github.com/roboflow/supervision</a></div>")
    #inp = gr.Image(type="filepath")
    with gr.Row():
        with gr.Column():
            inp = gr.Video()
            btn = gr.Button()
        outp_v = gr.Video(label="Processed Video", streaming=True, autoplay=True)
    outp_j = gr.JSON()

    btn.click(stream_object_detection,inp,[outp_v,outp_j])
app.queue().launch()