Sample Overview

Sample Obtaining

Click image decoding+cropping and resizing+image encoding+synchronous inference to obtain the sample.

Function Usage

This sample shows how to classify images based on the Caffe ResNet-50 network (single input with batch size = 1).

According to the arguments of the app, the following functions can be implemented:

Decode a .jpg image into an input image in YUV420SP (NV12) format, resize the image, and perform model inference. Then process the inference result and output the top 5 confidence values.
Crop an input image in YUV420SP (NV12) format and perform model inference. Then process the inference result and output the top 5 confidence values.
Crop and paste an input image in YUV420SP (NV12) format and perform model inference. Then process the inference result and output the top 5 confidence values.
Encode an input image in YUV420SP (NV12) format into a .jpg image.
Resize the 8192 x 8192 input image in YUV420SP (NV12) format to obtain a 4000 x 4000 subimage.
Crop the input images in YUV420SP (NV12) format in batches.
Crop and paste the input images in YUV420SP (NV12) format in batches.

Main APIs

The following table lists the main APIs.

Initialization	Call acl.init to initialize the configuration. Call aclFinalize for deinitialization.
Device management	Call acl.rt.set_device to specify the compute device. Call acl.rt.get_run_mode to obtain the running mode of the software stack. The internal processing process varies according to the running mode. Call acl.rt.reset_device to reset the current device and reclaim the associated resources.
Stream management	Call acl.rt.create_stream to create a stream. Call acl.rt.destroy_stream to destroy a stream. Call acl.rt.synchronize_stream to block the programs until all tasks in the specified stream are complete.
Memory management	Call acl.rt.malloc_host to allocate host memory. Call acl.rt.free_host to deallocate the memory on the host. Call acl.rt.malloc to allocate device memory. Call acl.rt.free to deallocate device memory. If device memory is needed to store the input or output data before media data processing, call acl.media.dvpp_malloc to allocate memory and acl.media.dvpp_free to deallocate memory.
Data transfer	If your app runs on the host, call the acl.rt.memcpy API. Transfers decode source data from the host to the device. Transfers the inference result from the device to the host. Data transfer is not required if your app runs in the board environment.
Media data processing V1	Image encoding Call acl.media.dvpp_jpeg_encode_async to encode a YUV420SP image into a .jpg image. Image decoding Call acl.media.dvpp_jpeg_decode_async* to decode .jpg images into YUV420SP images. Resizing Call acl.media.dvpp_vpc_resize_async* to resize the YUV420SP input. Cropping Call acl.media.dvpp_vpc_crop_resize_async to crop the specified area of the input image, and then overwrite the cropped area to the output buffer as the output image. Cropping and pasting Call acl.media.dvpp_vpc_crop_resize_paste_async to crop the specific area of the input image, and then overwrite the cropped area to the specified area of the target image as the output image.
Model inference	Call acl.mdl.load_from_file_with_mem to load a model from an .om file. Call acl.mdl.execute to perform model inference. Before inference, use the CSC parameters in the .om file to convert a YUV420SP image into an RGB image. Call acl.mdl.unload to unload a model.

Directory Structure

The directory structure is as follows:

vpc_jpeg_resnet50_imagenet_classification
├──scripts
│ ├── host_version.conf // Version number configuration file.
│ └── testcase_300.sh // Run script.
├──src
│ ├── acl_dvpp_process.py // Media data processing.
│ ├── acl_model_process.py // Model running file.
│ ├── constant.py // Constant definition.
│ └── main.py // Running file for process logic control.
├── data
│ ├── fusion_result.json // File generated after atc conversion, which records the fused operator information.
│ └── vdec_h265_1frame_rabbit_1280x720.h265 // Video file to be processed by the user, which is obtained by the user.
├── caffe_model // Model deployed by users.
│ ├── aipp.cfg // Model configuration data.
│ ├── resnet50.caffemodel // ResNet-50 model
│ └── resnet50.prototxt // ResNet-50 network file
└── model // Directory generated after the inference model is converted.
  └── resnet50_aipp.om // Model file generated after conversion.

Parent topic: Image Classification with Caffe ResNet-50 (Image Decoding+Cropping and Resizing+Image Encoding+Synchronous Inference)