§
    l¬¸jV	  ã                  ón   — d dl mZ d dlmZ d dlZd dlZd dlmZ d dl	m
Z
 ddlmZ  G d„ d	e¦  «        ZdS )
é    )Úannotations)ÚPathN)ÚLOGGER)Úcheck_requirementsé   )ÚBaseBackendc                  ó"   — e Zd ZdZdd„Zdd
„ZdS )ÚNCNNBackendzôTencent NCNN inference backend for mobile and embedded deployment.

    Loads and runs inference with Tencent NCNN models (*_ncnn_model/ directories). Optimized for mobile platforms with
    optional Vulkan GPU acceleration when available.
    Úweightú
str | PathÚreturnÚNonec                ó¦  — t          j        d|› d�¦  «         t          dd¬¦  «         ddl}|| _        |                     ¦   «         | _        t          | j        t          ¦  «        rŠ| j         
                    d¦  «        rpd	| j        j        _        | j                             t          | j                             d
¦  «        d         ¦  «        ¦  «         t!          j        d¦  «        | _        nd| j        j        _        t#          |¦  «        }|                     ¦   «         s"t'          |                     d¦  «        ¦  «        }| j                             t          |¦  «        ¦  «         | j                             t          |                     d¦  «        ¦  «        ¦  «         |                      |                      |¦  «        ¦  «         dS )z¾Load an NCNN model from a .param/.bin file pair or model directory.

        Args:
            weight (str | Path): Path to the .param file or directory containing NCNN model files.
        zLoading z for NCNN inference...Úncnnz	--no-deps)Úcmdsr   NÚvulkanTú:r   ÚcpuFz*.paramz.bin)r   Úinfor   r   ÚpyncnnÚNetÚnetÚ
isinstanceÚdeviceÚstrÚ
startswithÚoptÚuse_vulkan_computeÚset_vulkan_deviceÚintÚsplitÚtorchr   Úis_fileÚnextÚglobÚ
load_paramÚ
load_modelÚwith_suffixÚapply_metadataÚread_metadata)Úselfr   r   Úws       úl/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/ultralytics/nn/backends/ncnn.pyr'   zNCNNBackend.load_model   s‚  € õ 	ŒÐ=˜vÐ=Ð=Ð=Ñ>Ô>Ð>Ý˜6¨Ð4Ñ4Ô4Ð4ØÐÐÐàˆŒØ—:’:‘<”<ˆŒõ �d”k¥3Ñ'Ô'ð 	4¨D¬K×,BÒ,BÀ8Ñ,LÔ,Lð 	4Ø.2ˆDŒHŒLÔ+ØŒH×&Ò&¥s¨4¬;×+<Ò+<¸SÑ+AÔ+AÀ!Ô+DÑ'EÔ'EÑFÔFÐFÝœ, uÑ-Ô-ˆDŒKˆKà.3ˆDŒHŒLÔ+å�‰LŒLˆØ�yŠy‰{Œ{ð 	(Ý�Q—V’V˜IÑ&Ô&Ñ'Ô'ˆAàŒ×Ò�C ™FœFÑ#Ô#Ð#ØŒ×Ò�C §¢¨fÑ 5Ô 5Ñ6Ô6Ñ7Ô7Ð7à×Ò˜D×.Ò.¨qÑ1Ô1Ñ2Ô2Ð2Ð2Ð2ó    Úimútorch.Tensorúlist[np.ndarray]c           	     ó  ‡— g }|                      ¦   «                              ¦   «         D ]Ã}| j                             ¦   «         5 Š‰                     | j                             ¦   «         d         | j                             |¦  «        ¦  «         |                     ˆfd„t          | j         
                    ¦   «         ¦  «        D ¦   «         ¦  «         ddd¦  «         n# 1 swxY w Y   ŒÄd„ t          |Ž D ¦   «         S )a
  Run inference using the NCNN runtime.

        Args:
            im (torch.Tensor): Input image tensor in BCHW format, normalized to [0, 1].

        Returns:
            (list[np.ndarray]): Model predictions as a list of numpy arrays, one per output layer.
        r   c                ój   •— g | ]/}t          j        ‰                     |¦  «        d          ¦  «        ‘Œ0S )r   )ÚnpÚarrayÚextract)Ú.0ÚxÚexs     €r-   ú
<listcomp>z'NCNNBackend.forward.<locals>.<listcomp>C   s2   ø€ ÐdÐdÐd¸q¥¤¨¯ª°A©¬°qÔ)9Ñ :Ô :ÐdÐdÐdr.   Nc                ó6   — g | ]}t          j        |¦  «        ‘ŒS © )r4   Ústack)r7   Úys     r-   r:   z'NCNNBackend.forward.<locals>.<listcomp>D   s    € Ð3Ð3Ð3 •”˜‘”Ð3Ð3Ð3r.   )r   Únumpyr   Úcreate_extractorÚinputÚinput_namesr   ÚMatÚappendÚsortedÚoutput_namesÚzip)r+   r/   ÚoutputsÚsampler9   s       @r-   ÚforwardzNCNNBackend.forward5   s0  ø€ ð ˆØ—f’f‘h”h—n’nÑ&Ô&ð 	fð 	fˆFØ”×*Ò*Ñ,Ô,ð f°Ø—’˜œ×-Ò-Ñ/Ô/°Ô2°D´K·O²OÀFÑ4KÔ4KÑLÔLÐLà—’ÐdÐdÐdÐdÅFÈ4Ì8×K`ÒK`ÑKbÔKbÑDcÔDcÐdÑdÔdÑeÔeÐeðfð fð fñ fô fð fð fð fð fð fð føøøð fð fð fð føð 4Ð3¥S¨' ]Ð3Ñ3Ô3Ð3s   ÁBC"Ã"C&	Ã)C&	N)r   r   r   r   )r/   r0   r   r1   )Ú__name__Ú
__module__Ú__qualname__Ú__doc__r'   rJ   r<   r.   r-   r
   r
      sF   € € € € € ðð ð3ð 3ð 3ð 3ð<4ð 4ð 4ð 4ð 4ð 4r.   r
   )Ú
__future__r   Úpathlibr   r?   r4   r"   Úultralytics.utilsr   Úultralytics.utils.checksr   Úbaser   r
   r<   r.   r-   ú<module>rT      sª   ðð #Ð "Ð "Ð "Ð "Ð "à Ð Ð Ð Ð Ð à Ð Ð Ð Ø €€€à $Ð $Ð $Ð $Ð $Ð $Ø 7Ð 7Ð 7Ð 7Ð 7Ð 7à Ð Ð Ð Ð Ð ð44ð 44ð 44ð 44ð 44�+ñ 44ô 44ð 44ð 44ð 44r.   