blakeblackshear.frigate/frigate/process_clip.py

import datetime
import json
import logging
import multiprocessing as mp
import os
import subprocess as sp
import sys
from unittest import TestCase, main

import click
import cv2
import numpy as np

from frigate.config import FRIGATE_CONFIG_SCHEMA, FrigateConfig
from frigate.edgetpu import LocalObjectDetector
from frigate.motion import MotionDetector
from frigate.object_processing import COLOR_MAP, CameraState
from frigate.objects import ObjectTracker
from frigate.util import (DictFrameManager, EventsPerSecond,
                          SharedMemoryFrameManager, draw_box_with_label)
from frigate.video import (capture_frames, process_frames,
                           start_or_restart_ffmpeg)

logging.basicConfig()
logging.root.setLevel(logging.DEBUG)

logger = logging.getLogger(__name__)

def get_frame_shape(source):
    ffprobe_cmd = " ".join([
        'ffprobe',
        '-v',
        'panic',
        '-show_error',
        '-show_streams',
        '-of',
        'json',
        '"'+source+'"'
    ])
    p = sp.Popen(ffprobe_cmd, stdout=sp.PIPE, shell=True)
    (output, err) = p.communicate()
    p_status = p.wait()
    info = json.loads(output)

    video_info = [s for s in info['streams'] if s['codec_type'] == 'video'][0]

    if video_info['height'] != 0 and video_info['width'] != 0:
        return (video_info['height'], video_info['width'], 3)
    
    # fallback to using opencv if ffprobe didnt succeed
    video = cv2.VideoCapture(source)
    ret, frame = video.read()
    frame_shape = frame.shape
    video.release()
    return frame_shape

class ProcessClip():
    def __init__(self, clip_path, frame_shape, config: FrigateConfig):
        self.clip_path = clip_path
        self.camera_name = 'camera'
        self.config = config
        self.camera_config = self.config.cameras['camera']
        self.frame_shape = self.camera_config.frame_shape
        self.ffmpeg_cmd = [c['cmd'] for c in self.camera_config.ffmpeg_cmds if 'detect' in c['roles']][0]
        self.frame_manager = SharedMemoryFrameManager()
        self.frame_queue = mp.Queue()
        self.detected_objects_queue = mp.Queue()
        self.camera_state = CameraState(self.camera_name, config, self.frame_manager)

    def load_frames(self):
        fps = EventsPerSecond()
        skipped_fps = EventsPerSecond()
        current_frame = mp.Value('d', 0.0)
        frame_size = self.camera_config.frame_shape_yuv[0] * self.camera_config.frame_shape_yuv[1]
        ffmpeg_process = start_or_restart_ffmpeg(self.ffmpeg_cmd, logger, sp.DEVNULL, frame_size)
        capture_frames(ffmpeg_process, self.camera_name, self.camera_config.frame_shape_yuv, self.frame_manager, 
            self.frame_queue, fps, skipped_fps, current_frame)
        ffmpeg_process.wait()
        ffmpeg_process.communicate()
    
    def process_frames(self, objects_to_track=['person'], object_filters={}):
        mask = np.zeros((self.frame_shape[0], self.frame_shape[1], 1), np.uint8)
        mask[:] = 255
        motion_detector = MotionDetector(self.frame_shape, mask, self.camera_config.motion)

        object_detector = LocalObjectDetector(labels='/labelmap.txt')
        object_tracker = ObjectTracker(self.camera_config.detect)
        process_info = {
            'process_fps': mp.Value('d', 0.0),
            'detection_fps': mp.Value('d', 0.0),
            'detection_frame': mp.Value('d', 0.0)
        }
        stop_event = mp.Event()
        model_shape = (self.config.model.height, self.config.model.width)

        process_frames(self.camera_name, self.frame_queue, self.frame_shape, model_shape, 
            self.frame_manager, motion_detector, object_detector, object_tracker, 
            self.detected_objects_queue, process_info, 
            objects_to_track, object_filters, mask, stop_event, exit_on_empty=True)
    
    def top_object(self, debug_path=None):
        obj_detected = False
        top_computed_score = 0.0
        def handle_event(name, obj, frame_time):
            nonlocal obj_detected
            nonlocal top_computed_score
            if obj.computed_score > top_computed_score:
                top_computed_score = obj.computed_score
            if not obj.false_positive:
                obj_detected = True
        self.camera_state.on('new', handle_event)
        self.camera_state.on('update', handle_event)

        while(not self.detected_objects_queue.empty()):
            camera_name, frame_time, current_tracked_objects, motion_boxes, regions = self.detected_objects_queue.get()
            if not debug_path is None:
                self.save_debug_frame(debug_path, frame_time, current_tracked_objects.values())

            self.camera_state.update(frame_time, current_tracked_objects, motion_boxes, regions)
        
        self.frame_manager.delete(self.camera_state.previous_frame_id)
        
        return {
            'object_detected': obj_detected,
            'top_score': top_computed_score
        }
    
    def save_debug_frame(self, debug_path, frame_time, tracked_objects):
        current_frame = cv2.cvtColor(self.frame_manager.get(f"{self.camera_name}{frame_time}", self.camera_config.frame_shape_yuv), cv2.COLOR_YUV2BGR_I420)
        # draw the bounding boxes on the frame
        for obj in tracked_objects:
            thickness = 2
            color = (0,0,175)

            if obj['frame_time'] != frame_time:
                thickness = 1
                color = (255,0,0)
            else:
                color = (255,255,0)

            # draw the bounding boxes on the frame
            box = obj['box']
            draw_box_with_label(current_frame, box[0], box[1], box[2], box[3], obj['id'], f"{int(obj['score']*100)}% {int(obj['area'])}", thickness=thickness, color=color)
            # draw the regions on the frame
            region = obj['region']
            draw_box_with_label(current_frame, region[0], region[1], region[2], region[3], 'region', "", thickness=1, color=(0,255,0))
        
        cv2.imwrite(f"{os.path.join(debug_path, os.path.basename(self.clip_path))}.{int(frame_time*1000000)}.jpg", current_frame)

@click.command()
@click.option("-p", "--path", required=True, help="Path to clip or directory to test.")
@click.option("-l", "--label", default='person', help="Label name to detect.")
@click.option("-t", "--threshold", default=0.85, help="Threshold value for objects.")
@click.option("-s", "--scores", default=None, help="File to save csv of top scores")
@click.option("--debug-path", default=None, help="Path to output frames for debugging.")
def process(path, label, threshold, scores, debug_path):
    clips = []
    if os.path.isdir(path):
        files = os.listdir(path)
        files.sort()
        clips = [os.path.join(path, file) for file in files]
    elif os.path.isfile(path):  
        clips.append(path)

    json_config = {
        'mqtt': {
            'host': 'mqtt'
        },
        'cameras': {
            'camera': {
                'ffmpeg': {
                    'inputs': [
                        { 'path': 'path.mp4', 'global_args': '', 'input_args': '', 'roles': ['detect'] }
                    ]
                },
                'height': 1920,
                'width': 1080
            }
        }
    }

    results = []
    for c in clips:
        logger.info(c)
        frame_shape = get_frame_shape(c)
        
        json_config['cameras']['camera']['height'] = frame_shape[0]
        json_config['cameras']['camera']['width'] = frame_shape[1]
        json_config['cameras']['camera']['ffmpeg']['inputs'][0]['path'] = c

        config = FrigateConfig(config=FRIGATE_CONFIG_SCHEMA(json_config))

        process_clip = ProcessClip(c, frame_shape, config)
        process_clip.load_frames()
        process_clip.process_frames(objects_to_track=[label])

        results.append((c, process_clip.top_object(debug_path)))

    if not scores is None:
        with open(scores, 'w') as writer:
            for result in results:
                writer.write(f"{result[0]},{result[1]['top_score']}\n")
    
    positive_count = sum(1 for result in results if result[1]['object_detected'])
    print(f"Objects were detected in {positive_count}/{len(results)}({positive_count/len(results)*100:.2f}%) clip(s).")

if __name__ == '__main__':
    process()