o Ďi @sdZdgZddlZddlmZmZddlmZddlm Z ddl Z ddl Z ddlZddlZddlZddlmZddlmZdd lmZzdd lmZWneyZdd lmZYnwdd lmZdd lmZddl m!Z!ddl m"Z"GdddZ#dS)z%Detector functionality of SpeciesNet.SpeciesNetDetectorN)AnyOptional)logging)format_timespan) letterbox)non_max_suppression) xyxy2xywhn) scale_boxes) scale_coords) Detection)Failure) ModelInfo)PreprocessedImagec@seZdZdZdZdZdZdeddfdd Zd e e j j de e fd d Z d eedeefddZded e e deeeffddZdS)rz!Detector component of SpeciesNet.i@g{Gz? model_namereturnNcCs t}t||_tjrd|_n tjjrd|_nd|_|jdkr7tj |jj |jdd}|d |_ ntj |jj dd}|d |j|_ |j |j D]}d|_qU|j D]}t|tjjrqt|dsqd |_q`t}td t|||jd S) aLoads the detector resources. Code adapted from: https://github.com/agentmorris/MegaDetector which was released under the MIT License: https://github.com/agentmorris/MegaDetector/blob/main/LICENSE Args: model_name: String value identifying the model to be loaded. It can be a Kaggle identifier (starting with `kaggle:`), a HuggingFace identifier (starting with `hf:`) or a local folder to load the model from. cudampscpuF) map_location weights_onlymodel)rrecompute_scale_factorNz&Loaded SpeciesNetDetector in %s on %s.)timer model_infotorchr is_availabledevicebackendsrloaddetectorfloatrtoeval parameters requires_gradmodules isinstancennUpsamplehasattrrrinforupper)selfr start_time checkpointparammend_timer4O/home/jeff/fluffinator/venv/lib/python3.10/site-packages/speciesnet/detector.py__init__5s:       zSpeciesNetDetector.__init__imgcCs:|durdStt|tjtjddd}t||j|jS)aPreprocesses an image according to this detector's needs. Args: img: PIL image to preprocess. If `None`, no preprocessing is performed. Returns: A preprocessed image, or `None` if no PIL image was provided initially. NT) new_shapestrideautor) yolov5_letterboxnpasarrayrIMG_SIZESTRIDErwidthheight)r.r7img_arrr4r4r5 preprocessms zSpeciesNetDetector.preprocess yolo_xywhncCs0|\}}}}||d}||d}||||gS)agConverts bbox XYWHN coordinates from YOLO's to MegaDetector's format. Args: yolo_xywhn: List of bbox coordinates in YOLO format, i.e. [x_center, y_center, width, height]. Returns: List of bbox coordinates in MegaDetector format, i.e. [x_min, y_min, width, height]. g@r4)r.rDx_centery_centerr@rAx_miny_minr4r4r5_convert_yolo_xywhn_to_md_xywhns   z2SpeciesNetDetector._convert_yolo_xywhn_to_md_xywhnfilepathcCs|dur |tjjgdSt|jd}|gd}t|d}| |j }|j |ddd}|j dkr;| }t |tjd }|d}g}t|jd d|dddd f|j|jf|dddd f<|D]\}|dd }|j} | d kr|dddf}t||j|jd } | d kr| d} || } |d } tt|dd } t| }|durtd| qk| | |j!| | dqkt"|dddd}||dS)a(Runs inference on a given preprocessed image. Code adapted from: https://github.com/agentmorris/MegaDetector which was released under the MIT License: https://github.com/agentmorris/MegaDetector/blob/main/LICENSE Args: filepath: Location of image to run inference on. Used for reporting purposes only, and not for loading the image. img: Preprocessed image to run inference on. If `None`, a failure message is reported back. Returns: A dict containing either the detections above a fixed confidence threshold for the given image (in decreasing order of confidence scores), or a failure message if no preprocessed image was provided. N)rJfailures)rrF)augmentr) prediction conf_thresrMrN)whzInvalid detection class: %s)categorylabelconfbboxcSs|dS)NrXr4)detr4r4r5sz,SpeciesNetDetector.predict..T)keyreverse)rJ detections)#r DETECTORnamer from_numpyarrpermute unsqueezer"r#rrryolov5_non_max_suppressionrDETECTION_THRESHOLDyolov5_scale_boxesshape orig_height orig_widthroundndimyolov5_xyxy2xywhnrItolistitemstrintr from_categoryrerrorappendvaluesorted)r.rJr7 img_tensor batch_tensorresultsr^resultxyxyndimsxywhnrYrXrVrWr4r4r5predictsd         zSpeciesNetDetector.predict)__name__ __module__ __qualname____doc__r>r?rfrpr6rPILImagerrClistr"rIdictrr~r4r4r4r5r.s8 )$r__all__rtypingrrabslr humanfriendlyrnumpyr< PIL.Imagerrtorch.backendstorch.backends.mpsyolov5.utils.augmentationsrr;yolov5.utils.generalrrer rmr rg ImportErrorr speciesnet.constantsr r speciesnet.utilsrrrr4r4r4r5s2