Emmanuel Benazera

DeepDetect with Apple Silicon ARM M1 support via NCNN with Vulkan

10

This PR adds Vulkan support to DeepDetect for Apple ARM M1 with NCNN. At the moment: - PR relies on #1103 - Vulkan support is Apple ARM M1 only, will...

DeepDetect builds with `TENSORRT_OSS` do require CUDA 10.2

2

Otherwise the build error is as follows: ``` deepdetect/build/tensorrt-oss/src/tensorrt-oss/samples/common/sampleDevice.h:189:56: error: ‘cudaStreamCaptureModeGlobal’ was not declared in this scope cudaCheck(cudaStreamBeginCapture(stream.get(), cudaStreamCaptureModeGlobal)); ^~~~~~~~~~~~~~~~~~~~~~~~~~~ deepdetect/build/tensorrt-oss/src/tensorrt-oss/samples/common/sampleDevice.h:189:56: note: suggested alternative: ‘cudaStreamCaptureStatusNone’ cudaCheck(cudaStreamBeginCapture(stream.get(), cudaStreamCaptureModeGlobal)); ^~~~~~~~~~~~~~~~~~~~~~~~~~~ cudaStreamCaptureStatusNone ```...

This is tested on RTX 2080ti, `fp16` works, `int8` fails with the error below: ``` [error] Parameter check failed at: ../builder/Network.cpp::addScale::433, condition: scale.type == shift.type && shift.type == power.type error...

Ability to call `POST /predict` for images: - in classification, returns the classes - for object detection, returns the objects, classes + ability to screen the image with bounding boxes...

PR #706 greatly improved the docker images. However, we need a second pass to clean up / reorganize some elements. See the list below: - the `build` dir is often...

Recurrent neural layers support (RNN, LSTM) via Caffe backend (Direct inputs to LSTM)

43

RNN + LSTM support now merged into Caffe, https://github.com/BVLC/caffe/pull/3948. This paves the way for robust integration within dd.

See code and models here: https://github.com/tensorflow/models/tree/master/syntaxnet Input should be sentence and output a tree with predicted elements.

Unsafe use of Tensorflow and Caffe on same DeepDetect server / GPU

3

Tensorflow running call conflicts on Cuda initialization context with that of other libraries, see: - https://github.com/tensorflow/tensorflow/issues/916 - https://github.com/tensorflow/tensorflow/issues/2210 - https://github.com/mosessoh/CNN-LSTM-Caption-Generator/issues/2 - https://github.com/mdeff/cnn_graph/issues/1 Tensorflow devs indicate that there's no solving of...