import cv2 import copy import shutil import time import numpy as np from datetime import datetime from cached_property import cached_property from pathlib import Path from numpy import uint8, fromfile from rich.table import Table from threading import Event, Lock, Thread from oashya.labels import id2label, CLASSIFY, CLASSINDEX, id2name from oashya.utils import draw_tracks from oashya.tracker import Tracker from tasks.base_task import BaseTask from module.logger import logger from tasks.Hyakkiyakou.agent.focus import Focus def test_track(show: bool = False): tracker = Tracker() # -------------------------------- # folder_save = Path('./tasks/Hyakkiyakou/temp/save') shutil.rmtree(folder_save, ignore_errors=True) folder_save.mkdir(parents=True, exist_ok=True) folder_image = Path('./tasks/Hyakkiyakou/temp/20240621T221325') # start_time = datetime.now() for index, file in enumerate(folder_image.iterdir()): print(f'{index}-----------------------------------------------------------------------------------------------') if file.suffix != '.png': continue image = cv2.imdecode(fromfile(str(file), dtype=uint8), -1) image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) results = tracker(image=image, response=[0, 0, False, 10]) print(results) save_file = folder_save / f'{file.stem}.png' draw_image = draw_tracks(image, results) if show: cv2.imshow('OAS Tracker', draw_image) cv2.waitKey(500) print('Press any key to continue') else: cv2.imwrite(str(save_file), draw_image) # if index >= 14: # raise end_time = datetime.now() print(f'平均时间: {(end_time - start_time) / len([folder_image.iterdir()])}') def show_track(sync_image, sync_lock, stop_event): # global glob_image while True: if stop_event.is_set(): logger.info('OAS Track Debugger stopped') cv2.destroyAllWindows() break sync_lock.acquire() if Debugger.sync_image is None: logger.info('OAS Track Debugger not started') time.sleep(1) sync_lock.release() continue cv2.imshow('OAS Tracker', Debugger.sync_image) sync_lock.release() cv2.waitKey(100) class Debugger: info_enable: bool = False sync_image = None sync_lock = Lock() stop_event = Event() sync_thread = Thread(target=show_track, daemon=True, name='OAS Track Debugger', args=(sync_image, sync_lock, stop_event)) # sync_event = Event() # reserved for future use # ------------------------------------------------------------------------- def __init__(self, info_enable: bool = False, continuous_learning: bool = False, hya_save_result: bool = False): self._reset_thread_env() Debugger.info_enable = info_enable self.images_cache: dict = {} self.continuous_learning = continuous_learning self.hya_save_result = hya_save_result if continuous_learning: logger.info('Continuous Learning Mode Enabled') save_time = datetime.now().strftime('%Y%m%dT%H') self.hya_save_folder: Path = Path(f'./log/hya/{save_time}') self.hya_save_folder.mkdir(parents=True, exist_ok=True) if hya_save_result: logger.info('Hyakkiyakou Save Result Mode Enabled') save_time = datetime.now().strftime('%Y%m%dT%H') self.hya_save_result_folder: Path = Path(f'./log/hyakkiyakou/{save_time}') self.hya_save_result_folder.mkdir(parents=True, exist_ok=True) @cached_property def save_class(self) -> list[int]: """ 需要保存的类别 @return: """ sp = [i for i in range(CLASSINDEX.MIN_SP, CLASSINDEX.MAX_SP + 1)] ssr = [i for i in range(CLASSINDEX.MIN_SSR, CLASSINDEX.MAX_SSR + 1)] g = [i for i in range(CLASSINDEX.MIN_G, CLASSINDEX.MAX_G + 1)] return sp + ssr + g def _reset_thread_env(self): logger.info('Reset Debugger Thread Environment') Debugger.sync_image = None self.sync_lock = Lock() self.stop_event = Event() self.sync_thread = Thread(target=show_track, daemon=True, name='OAS Track Debugger', args=(Debugger.sync_image, self.sync_lock, self.stop_event)) def show(self, results): image = draw_tracks(self.device.image, results) cv2.imshow('OAS Tracker', image) def show_start(self): logger.info('OAS Track Debugger show image start') Debugger.sync_image = np.zeros((720, 1280, 3), dtype=np.uint8) self.sync_thread.start() def show_sync(self, image=None): self.sync_lock.acquire() if image is not None: # deepcopy Debugger.sync_image = copy.deepcopy(image) # logger.info('OAS Track Debugger synced image') self.sync_lock.release() def show_stop(self): logger.info('OAS Track Debugger show image stop') self.stop_event.set() self.sync_thread.join() def check_class(self, _class: int) -> bool: return _class in self.save_class def deal_learning(self, image, tracks: list): save_flag: bool = False for _id, _class, _conf, _cx, _cy, _w, _h, _v in tracks: if self.check_class(_class): save_flag = True break if not save_flag: return time_now_image_name = f'hya_{int(time.time() * 1000)}' self.images_cache[time_now_image_name] = image def save_images(self): if not self.images_cache: self.images_cache: dict = {} return logger.info('OAS Track Debugger save images to train model') for image_name, image in self.images_cache.items(): image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) cv2.imwrite(str(self.hya_save_folder / f'{image_name}.png'), image) self.images_cache.clear() def save_result(self, image): if not self.hya_save_result: return _now_name = f'hya_{int(time.time() * 1000)}.png' image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) cv2.imwrite(str(self.hya_save_result_folder / _now_name), image) @classmethod def show_info(cls, tracker, f: Focus): table = Table(show_lines=True) table.add_column('Property', header_style="bright_cyan", style="cyan", no_wrap=True) table.add_column("Value", style="green") table.add_column('Property', header_style="bright_cyan", style="cyan", no_wrap=True) table.add_column("Value", style="green") if f is not None: try: table.add_row('id', str(f._id), 'conf', f'{f._conf:.2f}') table.add_row('class', str(f._class), 'xywh', f'({f._cx}, {f._cy}, {f._w}, {f._h})') table.add_row('label', id2label(f._class), 'velocity', f'{f._v:.2f}') table.add_row('name', f'{id2name(f._class)}', 'omega', str(f._omega)) except Exception as e: pass table.add_row('detect cost', f'{(tracker.detect_time.total_seconds() * 1000):.2f}ms', 'track cost', f'{(tracker.track_time.total_seconds() * 1000):.2f}ms') logger.print(table, justify='center') def test_debugger(): debugger = Debugger(info_enable=False, continuous_learning=True) debugger.show_start() folder_image = Path('./tasks/Hyakkiyakou/temp/20240621T221325') for index, file in enumerate(folder_image.iterdir()): print(f'{index}-----------------------------------------------------------------------------------------------') if file.suffix != '.png': continue image = cv2.imdecode(fromfile(str(file), dtype=uint8), -1) image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) debugger.show_sync(image) tracks = [[1, 190, 0.9, 0, 0, 0, 0, 0]] debugger.deal_learning(image=image, tracks=tracks) time.sleep(0.5) if index >= 19: break debugger.show_stop() debugger.save_images() if __name__ == '__main__': # test_track() test_debugger() pass