U
|
P´æ`T ã @ s d dl Z d dlZd dlZd dlm mZ d dlmZ d dl m
|
Z
|
d dl Z d dlZd dl m Z d dlmZ G dd deZG dd d ZdS )
|
é N)Ú build_model)Ú Checkpointer)Úsimplify)ÚOperatorExportTypesc @ s$ e Zd Zdd Zdd Zdd ZdS )Ú
|
ReID_Modelc C s || _ t|| _d S ©N)ÚcfgÚDefaultPredictorÚ predictor)Úselfr © r úT/data/disk1/workspace/06_reid/01_fast_reid/02_fast_reid_inference/tools/predictor.pyÚ__init__ s zReID_Model.__init__c C s | |¡}|S )zù
|
|
Args:
|
original_image (np.ndarray): an image of shape (H, W, C) (in BGR order).
|
This is the format used by OpenCV.
|
|
Returns:
|
predictions (np.ndarray): normalized feature of the model.
|
)r
|
)r Zoriginal_imageÚ predictionsr r r Ú run_on_image s
|
zReID_Model.run_on_imagec C s | j ¡ }d S r )r
|
Úto_onnx)r r r r r Ú
|
torch2onnx2 s zReID_Model.torch2onnxN)Ú__name__Ú
|
__module__Ú __qualname__r r r r r r r r s r c @ s8 e Zd ZdZdd Zdd Zdd Zdd Zd
|
d Zd S ) r aB
|
Create a simple end-to-end predictor with the given config.
|
The predictor takes an BGR image, resizes it to the specified resolution,
|
runs the model and produces a dict of predictions.
|
This predictor takes care of model loading and input preprocessing for you.
|
If you'd like to do anything more fancy, please refer to its source code
|
as examples to build and use the model manually.
|
Attributes:
|
Examples:
|
.. code-block:: python
|
pred = DefaultPredictor(cfg)
|
inputs = cv2.imread("input.jpg")
|
outputs = pred(inputs)
|
c C sn | ¡ | _| j ¡ d| jjj_t| j| _| j ¡ D ]
|
}d|_ q6| j
|
¡ | j ¡ t | j |jj¡ d S )NF)Úcloner ÚdefrostÚMODELÚBACKBONEÚPRETRAINr ÚmodelÚ
|
parametersÚ requires_gradÚcudaÚevalr ÚloadZWEIGHTS)r r Úparamr r r r E s
|
|
|
|
zDefaultPredictor.__init__c
|
C sj t ¡ X | ¡ }| j ¡ | |¡}t jdd td| t |¡}| ¡ j
|
}|W 5 Q R £ S Q R X dS )zº
|
Args:
|
image (torch.tensor): an image tensor of shape (B, C, H, W).
|
Returns:
|
predictions (torch.tensor): the output features of the model
|
i )Ú edgeitemsz ------------
|
N) ÚtorchÚno_gradr r r Úset_printoptionsÚprintÚFÚ normalizeÚcpuÚdata)r ÚimageZimagesr Z pred_featr r r Ú__call__Q s
|
|
|
|
|
|
zDefaultPredictor.__call__c C sh t dd| jjjd | jjjd ¡ ¡ }| | j|¡}t|\}}| |¡}|sXt
|
dt |d¡ d S )Né é r z,Simplified ONNX model could not be validatedz fastreid.onnx) r# Úrandnr ZINPUTZ SIZE_TESTr Úexport_onnx_modelr r Ú remove_initializer_from_inputÚAssertionErrorÚonnxZ
|
save_model)r ÚinputsÚ
|
onnx_modelZ
|
model_simpÚcheckr r r r h s (
|
zDefaultPredictor.to_onnxc
|
s° t |tjjstdd }| |¡ t ¡ > t ¡ *}tj j
|
|||t j d t | ¡ ¡}W 5 Q R X W 5 Q R X t j ¡ dddg}t fdd|D stt j ||¡}|S ) zâ
|
Trace and export a model to onnx format.
|
Args:
|
model (nn.Module):
|
inputs (torch.Tensor): the model will be called by `model(*inputs)`
|
Returns:
|
an onnx model
|
c S s | j r
|
td S r )Útrainingr2 )Úmoduler r r Ú _check_eval s z7DefaultPredictor.export_onnx_model.<locals>._check_eval)Úoperator_export_typeZ extract_constant_to_initializerZ eliminate_unused_initializerZfuse_bn_into_convc 3 s | ]}| kV qd S r r )Ú.0Úp©Z
|
all_passesr r Ú <genexpr> s z5DefaultPredictor.export_onnx_model.<locals>.<genexpr>)Ú
|
isinstancer# ÚnnÚModuler2 Úapplyr$ ÚioÚBytesIOr3 Úexportr ÚONNX_ATEN_FALLBACKZload_from_stringÚgetvalueÚ optimizerZget_available_passesÚallÚoptimize)r r r4 r9 Úfr5 Zpassesr r= r r0 y s"
|
|
|
ü"
|
|
z"DefaultPredictor.export_onnx_modelc C sb |j dk rtd d S |jj}i }|D ]}|||j< q&|jjD ] }|j|kr>| ||j ¡ q>|S )Né zKModel with ir_version below 4 requires to include initilizer in graph input)Ú
|
ir_versionr& ÚgraphÚinputÚnameÚ initializerÚremove)r r r4 Z name_to_inputrO rQ r r r r1 s
|
ÿ
|
z.DefaultPredictor.remove_initializer_from_inputN) r r r Ú__doc__r r, r r0 r1 r r r r r 5 s 'r )Zcv2rC r# Ztorch.nn.functionalr@ Ú
|
functionalr' Zmodeling.meta_arch.buildr Zutils.checkpointr r3 Zonnxsimr Ú
|
torch.onnxr Úobjectr r r r r r Ú<module> s "
|
