From d82673332cc3930675084612fe2760d147bdaeec Mon Sep 17 00:00:00 2001 From: Henry Ruhs Date: Sat, 8 Aug 2026 13:28:54 +0200 Subject: [PATCH] Feat/asset tools part2 (#1218) * add face tools to assets * add face tools to assets * is_vision_frames returns false on empty list --- facefusion/apis/asset_helper.py | 17 +++++++++++++++++ facefusion/apis/endpoints/assets.py | 12 +++++++++--- facefusion/types.py | 2 +- facefusion/vision.py | 2 +- 4 files changed, 28 insertions(+), 5 deletions(-) diff --git a/facefusion/apis/asset_helper.py b/facefusion/apis/asset_helper.py index c4766a7b..81544f86 100644 --- a/facefusion/apis/asset_helper.py +++ b/facefusion/apis/asset_helper.py @@ -7,6 +7,8 @@ from starlette.datastructures import UploadFile import facefusion.choices from facefusion import ffmpeg, process_manager, state_manager +from facefusion.face_creator import get_many_faces +from facefusion.face_helper import warp_face_by_bounding_box from facefusion.filesystem import create_directory, get_file_extension, get_file_format, is_audio, is_image, is_video from facefusion.types import ImageAsset, ImageMetadata, MediaType, VideoAsset, VisionFrame from facefusion.vision import detect_image_resolution, fit_contain_frame, is_vision_frame, read_static_image, read_static_video_frame, unpack_resolution @@ -120,3 +122,18 @@ def capture_asset_frames(asset : ImageAsset | VideoAsset, frame_numbers : List[s capture_vision_frames.append(capture_vision_frame) return capture_vision_frames + + +def capture_asset_faces(asset : ImageAsset | VideoAsset, frame_numbers : List[str], resolution : str) -> List[VisionFrame]: + capture_vision_frames = [] + temp_vision_frames = read_asset_frames(asset, frame_numbers) + crop_size = unpack_resolution(resolution) + + for temp_vision_frame in temp_vision_frames: + faces = get_many_faces([ temp_vision_frame ]) + + for face in faces: + capture_vision_frame, _ = warp_face_by_bounding_box(temp_vision_frame, face.bounding_box, crop_size) + capture_vision_frames.append(capture_vision_frame) + + return capture_vision_frames diff --git a/facefusion/apis/endpoints/assets.py b/facefusion/apis/endpoints/assets.py index 0e267efd..a56e801c 100644 --- a/facefusion/apis/endpoints/assets.py +++ b/facefusion/apis/endpoints/assets.py @@ -7,7 +7,7 @@ from starlette.status import HTTP_200_OK, HTTP_201_CREATED, HTTP_400_BAD_REQUEST from facefusion import session_context, session_manager from facefusion.apis import asset_store -from facefusion.apis.asset_helper import capture_asset_frames, save_asset_files, validate_asset_files +from facefusion.apis.asset_helper import capture_asset_faces, capture_asset_frames, save_asset_files, validate_asset_files from facefusion.apis.session_helper import extract_access_token from facefusion.filesystem import remove_file from facefusion.vision import is_vision_frames, to_strip_buffer @@ -90,10 +90,16 @@ async def get_asset(request : Request) -> Response: asset = asset_store.get_asset(session_id, asset_id) if asset: - if asset.get('media') in [ 'image', 'video' ] and request.query_params.get('action') == 'capture' and request.query_params.get('subject') == 'frame': + if asset.get('media') in [ 'image', 'video' ] and request.query_params.get('action') == 'capture': resolution = request.query_params.get('resolution') frame_numbers = request.query_params.getlist('frame_number') - vision_frames = capture_asset_frames(asset, frame_numbers, resolution) #type:ignore[arg-type] + vision_frames = [] + + if request.query_params.get('subject') == 'frame': + vision_frames = capture_asset_frames(asset, frame_numbers, resolution) #type:ignore[arg-type] + + if request.query_params.get('subject') == 'face': + vision_frames = capture_asset_faces(asset, frame_numbers, resolution) #type:ignore[arg-type] if is_vision_frames(vision_frames): return Response(content = to_strip_buffer(vision_frames), media_type = 'image/jpeg') diff --git a/facefusion/types.py b/facefusion/types.py index dc973005..db6ae03c 100755 --- a/facefusion/types.py +++ b/facefusion/types.py @@ -315,7 +315,7 @@ AssetSet : TypeAlias = Dict[AssetId, AudioAsset | ImageAsset | VideoAsset] AssetStore : TypeAlias = Dict[SessionId, AssetSet] AssetAction = Literal['capture'] -AssetSubject = Literal['frame'] +AssetSubject = Literal['frame', 'face'] BenchmarkMode = Literal['warm', 'cold'] BenchmarkResolution = Literal['240p', '360p', '540p', '720p', '1080p', '1440p', '2160p'] diff --git a/facefusion/vision.py b/facefusion/vision.py index 24ebb598..ade01d6e 100644 --- a/facefusion/vision.py +++ b/facefusion/vision.py @@ -211,7 +211,7 @@ def is_vision_frames(vision_frames : List[VisionFrame]) -> bool: if not is_vision_frame(vision_frame): return False - return True + return bool(vision_frames) def restrict_frame(vision_frame : VisionFrame, resolution : Resolution) -> VisionFrame: