diff --git a/.gitignore b/.gitignore index 442a702..9c496eb 100644 --- a/.gitignore +++ b/.gitignore @@ -360,4 +360,5 @@ $RECYCLE.BIN/ /test.py /test2.py /subtitle.ini -*_no_sub.mp4 \ No newline at end of file +*_no_sub.mp4 +/backend/ffmpeg/win_x64/ffmpeg.exe diff --git a/backend/config.py b/backend/config.py index 5bc8144..bc011a1 100644 --- a/backend/config.py +++ b/backend/config.py @@ -3,12 +3,14 @@ warnings.filterwarnings("ignore") import os import torch import logging +import platform +import stat from fsplit.filesplit import Filesplit logging.disable(logging.DEBUG) # 关闭DEBUG日志的打印 logging.disable(logging.WARNING) # 关闭WARNING日志的打印 device = "cuda" if torch.cuda.is_available() else "cpu" -BASE_DIR = os.path.dirname(__file__) +BASE_DIR = os.path.dirname(os.path.abspath(__file__)) LAMA_CONFIG = os.path.join(BASE_DIR, 'inpaint', 'lama', 'configs', 'prediction', 'default.yaml') LAMA_MODEL_PATH = os.path.join(BASE_DIR, 'models', 'big-lama') MODEL_VERSION = 'V4' @@ -22,6 +24,23 @@ if 'best.ckpt' not in (os.listdir(os.path.join(LAMA_MODEL_PATH, 'models'))): fs = Filesplit() fs.merge(input_dir=os.path.join(LAMA_MODEL_PATH, 'models')) -if 'inference.pdiparams' not in DET_MODEL_PATH: +if 'inference.pdiparams' not in os.listdir(DET_MODEL_PATH): fs = Filesplit() fs.merge(input_dir=DET_MODEL_PATH) + + +# 指定ffmpeg可执行程序路径 +sys_str = platform.system() +if sys_str == "Windows": + ffmpeg_bin = os.path.join('win_x64', 'ffmpeg.exe') +elif sys_str == "Linux": + ffmpeg_bin = os.path.join('linux_x64', 'ffmpeg') +else: + ffmpeg_bin = os.path.join('macos', 'ffmpeg') +FFMPEG_PATH = os.path.join(BASE_DIR, '', 'ffmpeg', ffmpeg_bin) + +if 'ffmpeg.exe' not in os.listdir(os.path.join(BASE_DIR, '', 'ffmpeg', 'win_x64')): + fs = Filesplit() + fs.merge(input_dir=os.path.join(BASE_DIR, '', 'ffmpeg', 'win_x64')) +# 将ffmpeg添加可执行权限 +os.chmod(FFMPEG_PATH, stat.S_IRWXU+stat.S_IRWXG+stat.S_IRWXO) diff --git a/backend/ffmpeg/linux_x64/ffmpeg b/backend/ffmpeg/linux_x64/ffmpeg new file mode 100755 index 0000000..13e56c8 Binary files /dev/null and b/backend/ffmpeg/linux_x64/ffmpeg differ diff --git a/backend/ffmpeg/macos/ffmpeg b/backend/ffmpeg/macos/ffmpeg new file mode 100644 index 0000000..b8c4347 Binary files /dev/null and b/backend/ffmpeg/macos/ffmpeg differ diff --git a/backend/ffmpeg/win_x64/ffmpeg_1.exe b/backend/ffmpeg/win_x64/ffmpeg_1.exe new file mode 100644 index 0000000..5b75d5f Binary files /dev/null and b/backend/ffmpeg/win_x64/ffmpeg_1.exe differ diff --git a/backend/ffmpeg/win_x64/ffmpeg_2.exe b/backend/ffmpeg/win_x64/ffmpeg_2.exe new file mode 100644 index 0000000..902df3f Binary files /dev/null and b/backend/ffmpeg/win_x64/ffmpeg_2.exe differ diff --git a/backend/ffmpeg/win_x64/ffmpeg_3.exe b/backend/ffmpeg/win_x64/ffmpeg_3.exe new file mode 100644 index 0000000..c0ff50e Binary files /dev/null and b/backend/ffmpeg/win_x64/ffmpeg_3.exe differ diff --git a/backend/ffmpeg/win_x64/fs_manifest.csv b/backend/ffmpeg/win_x64/fs_manifest.csv new file mode 100644 index 0000000..501738d --- /dev/null +++ b/backend/ffmpeg/win_x64/fs_manifest.csv @@ -0,0 +1,4 @@ +filename,filesize,encoding,header +ffmpeg_1.exe,50000000,, +ffmpeg_2.exe,50000000,, +ffmpeg_3.exe,13721856,, diff --git a/backend/main.py b/backend/main.py index 71d502c..3865858 100644 --- a/backend/main.py +++ b/backend/main.py @@ -1,3 +1,6 @@ +import shutil +import subprocess +import random import config import os from pathlib import Path @@ -8,6 +11,7 @@ sys.path.insert(0, os.path.dirname(os.path.abspath(__file__))) sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__)))) import importlib import numpy as np +import tempfile from tqdm import tqdm from tools.infer import utility from tools.infer.predict_det import TextDetector @@ -93,6 +97,7 @@ class SubtitleRemover: importlib.reload(config) # 线程锁 self.lock = threading.RLock() + uln = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789' # 用户指定的字幕区域位置 self.sub_area = sub_area # 视频路径 @@ -111,8 +116,10 @@ class SubtitleRemover: # 创建字幕检测对象 self.sub_detector = SubtitleDetect(self.video_path, self.sub_area) # 创建视频写对象 + self.video_temp_out_name = os.path.join(os.path.dirname(self.video_path), f'{self.vd_name}_{"".join(random.sample(uln, 8))}.mp4') + self.video_writer = cv2.VideoWriter(self.video_temp_out_name, cv2.VideoWriter_fourcc(*'mp4v'), self.fps, self.size) self.video_out_name = os.path.join(os.path.dirname(self.video_path), f'{self.vd_name}_no_sub.mp4') - self.video_writer = cv2.VideoWriter(self.video_out_name, cv2.VideoWriter_fourcc(*'mp4v'), self.fps, self.size) + @staticmethod def get_coordinates(dt_box): @@ -153,8 +160,12 @@ class SubtitleRemover: tbar.update(1) self.video_cap.release() self.video_writer.release() + # 将原音频合并到新生成的视频文件中 + self.merge_audio_to_video() + print(f"视频生字幕去除成功，文件路径：{self.video_out_name}") - def inpaint(self, img, mask): + @staticmethod + def inpaint( img, mask): img_inpainted = inpaint_img_with_lama(img, mask, config.LAMA_CONFIG, config.LAMA_MODEL_PATH, device=config.device) return img_inpainted @@ -177,6 +188,26 @@ class SubtitleRemover: masks.append(mask) return masks + def merge_audio_to_video(self): + temp = tempfile.NamedTemporaryFile(suffix='.aac', delete=False) + audio_extract_command = [config.FFMPEG_PATH, + "-y", "-i", self.video_path, + "-acodec", "copy", + "-vn", "-loglevel", "error", temp.name] + use_shell = True if os.name == "nt" else False + subprocess.check_output(audio_extract_command, stdin=open(os.devnull), shell=use_shell) + if os.path.exists(self.video_temp_out_name): + audio_merge_command = [config.FFMPEG_PATH, + "-y", "-i", self.video_temp_out_name, + "-i", temp.name, + "-vcodec", "copy", + "-acodec", "copy", + "-loglevel", "error", self.video_out_name] + subprocess.check_output(audio_merge_command, stdin=open(os.devnull), shell=use_shell) + if os.path.exists(self.video_temp_out_name): + os.remove(self.video_temp_out_name) + temp.close() + if __name__ == '__main__': # 提示用户输入视频路径 @@ -184,5 +215,3 @@ if __name__ == '__main__': # 新建字幕提取对象 sd = SubtitleRemover(video_path) sd.run() - -