o U6i @sdZddlZddlmZddlmZddlmZddlm Z de eBd e fd d Z dde eBd e eBde d efddZde eBd e fddZdS)z4Face and mouth region detection using Gemini Vision.N)Path) get_client) GEMINI_MODEL)types image_pathreturnc Cst}t|}t|d }|}Wdn1swYtjj|d|jddd}d}|j j t ||gd}|j }|dr^|dd}|d rZ|d d}| }zt|}d |t|d WStjy} zd d| |j t|dWYd} ~ Sd} ~ ww)z Detect face and mouth regions in an image using Gemini Vision. Args: image_path: Path to the image file Returns: dict with detected regions (normalized coordinates 0-1) rbNzimage/r)dataZ mime_typeaEAnalyze this face image and identify the regions. Return a JSON object with the following structure: { "face": { "x": , "y": , "width": , "height": }, "mouth": { "x": , "y": , "width": , "height": }, "eyes": { "left": {"x":
, "y":
}, "right": {"x":
, "y":
} }, "face_angle": "frontal" | "left_profile" | "right_profile" | "three_quarter_left" | "three_quarter_right", "art_style": "realistic" | "anime" | "cartoon" | "pixel" | "other", "confidence": <0-1> } All coordinates should be normalized between 0 and 1, where (0,0) is top-left. Only return the JSON, no other text.)Zmodelcontentsz```jsonT)successregionsrFzFailed to parse regions: )rerrorZ raw_responser)rropenreadrZPart from_bytessuffixlowerZmodelsZgenerate_contentrtextstrip startswithsplitr loadsstrZJSONDecodeError) rZclientfZ image_dataZ image_partpromptZresponserrerP/var/www/tkim.planitai.co.jp/blog/20251208-make-apng-tool/./src/face_detector.pydetect_face_regions sD        r!皙? output_pathpaddingcCsdddlm}t|}t|}t|}|ds td|d|d}|di}|s0td||}|j\}} |d d } |d d } |d d} |dd} td| || } td| || } t d| | dd|} t d| | dd|} t | |}t | | }t | | |}t | | | }| ||||f}|j j ddd|||S)a Crop the mouth region from an image. Args: image_path: Path to the source image output_path: Path to save the cropped image padding: Extra padding around the mouth region (0-1) Returns: Path to the cropped image r)ImagerzFailed to detect regions: rrmouthzNo mouth region detectedxg333333?yg333333?widthg?heightr"rT)parentsexist_ok)PILr%rr! ValueErrorgetrsizemaxminintcropparentmkdirsave)rr#r$r%resultrr&imgr)r*r'r(whlefttoprightbottomZcroppedrrr crop_mouth_regionXs8           rAcCst|}|dsdddgdS|d}g}|dd}|d vr$|d |d d }|d kr3|d|di}|rC|dd dkrH|dt|d kpQ|dk|||dS)z Validate that an image is suitable for lip sync generation. Args: image_path: Path to the image file Returns: dict with validation results and recommendations rFzCould not detect face regionsz'Use a clearer image with a visible face)validreasonrecommendationsr face_angleunknown)ZfrontalZthree_quarter_leftZthree_quarter_rightz5Use a frontal or 3/4 view face image for best results confidencergffffff?z3Image quality or face visibility may affect resultsr&r)g?z(Mouth region is too small or not visibleg?)rBrrDrG)r!r0appendlen)rr9rrDrErGr&rrr validate_face_images,       rJ)r")__doc__r pathlibrZ gemini_clientrZconfigrZ google.genairrdictr!floatrArJrrrr s    $O: