| """ |
| generate_annotations.py |
| ======================= |
| Gera o arquivo annotations.csv para o dataset minds-libras. |
| |
| Uso: |
| python generate_annotations.py |
| |
| O script deve ser colocado na pasta PAI da pasta 'videos', ou seja: |
| dataset/minds-libras/ |
| generate_annotations.py <-- aqui |
| annotations.csv <-- gerado aqui |
| videos/ |
| 01AcontecerSinalizador01-1.mp4 |
| ... |
| |
| Dependências: |
| - ffprobe (parte do ffmpeg): extrai width, height e fps reais de cada vídeo. |
| Caso contrário, as colunas ficam como None. |
| |
| Ubuntu/Debian: sudo apt install ffmpeg |
| macOS: brew install ffmpeg |
| """ |
|
|
| import os |
| import re |
| import csv |
| import json |
| import subprocess |
| from pathlib import Path |
|
|
| |
| |
| |
|
|
| SCRIPT_DIR = Path(__file__).parent.resolve() |
| VIDEOS_DIR = SCRIPT_DIR / "videos" |
| OUTPUT_CSV = SCRIPT_DIR / "annotations.csv" |
|
|
| FILENAME_PATTERN = re.compile( |
| r"^(?P<class_id>\d{2})(?P<class_name>[A-Za-záéíóúàâêôãõüçÁÉÍÓÚÀÂÊÔÃÕÜÇ]+)" |
| r"Sinalizador(?P<user_id>\d{2})-(?P<rep>\d+)\.mp4$" |
| ) |
|
|
| def get_video_metadata(filepath: Path) -> dict: |
| """Retorna width, height e frame_count via ffprobe. Retorna None em caso de erro.""" |
| try: |
| cmd = [ |
| "ffprobe", "-v", "quiet", |
| "-print_format", "json", |
| "-select_streams", "v:0", |
| "-show_streams", |
| str(filepath) |
| ] |
| result = subprocess.run(cmd, capture_output=True, text=True, timeout=10) |
| if result.returncode != 0: |
| return {"width": None, "height": None, "fps": None} |
|
|
| data = json.loads(result.stdout) |
| stream = data.get("streams", [{}])[0] |
|
|
| width = stream.get("width") |
| height = stream.get("height") |
|
|
| |
| nb_frames = stream.get("nb_frames") |
| if nb_frames and nb_frames != "N/A": |
| fps = float(nb_frames) |
| else: |
| |
| avg_fr = stream.get("avg_frame_rate", "0/1") |
| num, den = avg_fr.split("/") |
| fps = round(float(num) / float(den), 2) if float(den) != 0 else None |
|
|
| return {"width": width, "height": height, "fps": fps} |
|
|
| except (FileNotFoundError, subprocess.TimeoutExpired, Exception): |
| return {"width": None, "height": None, "fps": None} |
|
|
|
|
| def ffprobe_available() -> bool: |
| try: |
| subprocess.run(["ffprobe", "-version"], capture_output=True, timeout=5) |
| return True |
| except (FileNotFoundError, subprocess.TimeoutExpired): |
| return False |
|
|
| def main(): |
| if not VIDEOS_DIR.exists(): |
| raise FileNotFoundError(f"Pasta de vídeos não encontrada: {VIDEOS_DIR}") |
|
|
| use_ffprobe = ffprobe_available() |
| if use_ffprobe: |
| print("ffprobe detectado — extraindo metadados reais dos vídeos.") |
| else: |
| print("ffprobe NÃO encontrado — colunas width/height/fps serão None.") |
| print(" Para extrair metadados: sudo apt install ffmpeg (ou brew install ffmpeg)") |
|
|
| |
| mp4_files = sorted(VIDEOS_DIR.glob("*.mp4")) |
| print(f"\nArquivos .mp4 encontrados: {len(mp4_files)}") |
|
|
| rows = [] |
| skipped = [] |
|
|
| for filepath in mp4_files: |
| filename = filepath.name |
| m = FILENAME_PATTERN.match(filename) |
|
|
| if not m: |
| skipped.append(filename) |
| continue |
|
|
| class_name = m.group("class_name") |
| user_id = f"Sinalizador{m.group('user_id')}" |
| rep = m.group("rep") |
|
|
| video_id = filename |
|
|
|
|
| video_name = f"{class_name}_{user_id}-{rep}.mp4" |
|
|
| |
| if use_ffprobe: |
| meta = get_video_metadata(filepath) |
| print(f" {filename} -> {meta}") |
| else: |
| meta = {"width": None, "height": None, "fps": None} |
|
|
| rows.append({ |
| "video_id": video_id, |
| "video_name": video_name, |
| "class": class_name, |
| "user_id": user_id, |
| "width": meta["width"], |
| "height": meta["height"], |
| "fps": meta["fps"], |
| }) |
|
|
| |
| if skipped: |
| print(f"\n⚠️ {len(skipped)} arquivo(s) não reconhecido(s) pelo padrão e ignorado(s):") |
| for s in skipped: |
| print(f" {s}") |
|
|
| |
| fieldnames = ["video_id", "video_name", "class", "user_id", "width", "height", "fps"] |
|
|
| with open(OUTPUT_CSV, "w", newline="", encoding="utf-8") as f: |
| writer = csv.DictWriter(f, fieldnames=fieldnames) |
| writer.writeheader() |
| writer.writerows(rows) |
|
|
| |
| classes = sorted(set(r["class"] for r in rows)) |
| users = sorted(set(r["user_id"] for r in rows)) |
| reps = sorted(set( |
| re.search(r"-(\d+)\.mp4$", r["video_id"]).group(1) for r in rows |
| )) |
|
|
| print(f"\n✅ annotations.csv gerado em: {OUTPUT_CSV}") |
| print(f" Total de linhas (vídeos): {len(rows)}") |
| print(f" Classes ({len(classes)}): {classes}") |
| print(f" Sinalizadores ({len(users)}): {users}") |
| print(f" Repetições: {reps}") |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
