tkDNN is a Deep Neural Network library built with cuDNN and tensorRT primitives, specifically thought to work on NVIDIA Jetson Boards. It has been tested on TK1(branch cudnn2), TX1, TX2, AGX Xavier, Nano and several discrete GPUs. The main goal of this project is to exploit NVIDIA boards as much as possible to obtain the best inference performance. It does not allow training.

If you use tkDNN in your research, please cite the following paper. For use in commercial solutions, write at [email protected] and [email protected] or refer to https://hipert.unimore.it/ .

@inproceedings{verucchi2020systematic,
  title={A Systematic Assessment of Embedded Neural Networks for Object Detection},
  author={Verucchi, Micaela and Brilli, Gianluca and Sapienza, Davide and Verasani, Mattia and Arena, Marco and Gatti, Francesco and Capotondi, Alessandro and Cavicchioli, Roberto and Bertogna, Marko and Solieri, Marco},
  booktitle={2020 25th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA)},
  volume={1},
  pages={937--944},
  year={2020},
  organization={IEEE}
}

Inference FPS of yolov4 with tkDNN, average of 1200 images with the same dimension as the input size, on

• RTX 2080Ti (CUDA 10.2, TensorRT 7.0.0, Cudnn 7.6.5);
• Xavier AGX, Jetpack 4.3 (CUDA 10.0, CUDNN 7.6.3, tensorrt 6.0.1 );
• Xavier NX, Jetpack 4.4 (CUDA 10.2, CUDNN 8.0.0, tensorrt 7.1.0 ).
• Tx2, Jetpack 4.2 (CUDA 10.0, CUDNN 7.3.1, tensorrt 5.0.6 );
• Jetson Nano, Jetpack 4.4 (CUDA 10.2, CUDNN 8.0.0, tensorrt 7.1.0 ).

Results for COCO val 2017 (5k images), on RTX 2080Ti, with conf threshold=0.001

• tkDNN
  • Index
  • Dependencies
  • About OpenCV
  • How to compile this repo
  • Workflow
  • How to export weights
    • 1)Export weights from darknet
    • 2)Export weights for DLA34 and ResNet101
    • 3)Export weights for CenterNet
    • 4)Export weights for MobileNetSSD
  • Run the demo
    • FP16 inference
    • INT8 inference
  • mAP demo
  • Existing tests and supported networks
  • References
  • tkDNN on Windows 10 (experimental)
    • Dependencies-Windows
    • Compiling tkDNN on Windows
    • Run the demo on Windows
      • FP16 inference windows
      • INT8 inference windows
    • Known issues with tkDNN on Windows

This branch works on every NVIDIA GPU that supports the dependencies:

• CUDA 10.0
• CUDNN 7.603
• TENSORRT 6.01
• OPENCV 3.4
• yaml-cpp 0.5.2 (sudo apt install libyaml-cpp-dev)

To compile and install OpenCV4 with contrib us the script install_OpenCV4.sh. It will download and compile OpenCV in Download folder.

bash scripts/install_OpenCV4.sh

When using openCV not compiled with contrib, comment the definition of OPENCV_CUDACONTRIBCONTRIB in include/tkDNN/DetectionNN.h. When commented, the preprocessing of the networks is computed on the CPU, otherwise on the GPU. In the latter case some milliseconds are saved in the end-to-end latency.

Build with cmake. If using Ubuntu 18.04 a new version of cmake is needed (3.15 or above).

git clone https://github.com/ceccocats/tkDNN
cd tkDNN
mkdir build
cd build
cmake .. 
make

Steps needed to do inference on tkDNN with a custom neural network.

• Build and train a NN model with your favorite framework.
• Export weights and bias for each layer and save them in a binary file (one for layer).
• Export outputs for each layer and save them in a binary file (one for layer).
• Create a new test and define the network, layer by layer using the weights extracted and the output to check the results.
• Do inference.

Weights are essential for any network to run inference. For each test a folder organized as follow is needed (in the build folder):

    test_nn
        |---- layers/ (folder containing a binary file for each layer with the corresponding wieghts and bias)
        |---- debug/  (folder containing a binary file for each layer with the corresponding outputs)

Therefore, once the weights have been exported, the folders layers and debug should be placed in the corresponding test.

To export weights for NNs that are defined in darknet framework, use this fork of darknet and follow these steps to obtain a correct debug and layers folder, ready for tkDNN.

git clone https://git.hipert.unimore.it/fgatti/darknet.git
cd darknet
make
mkdir layers debug
./darknet export <path-to-cfg-file> <path-to-weights> layers

N.b. Use compilation with CPU (leave GPU=0 in Makefile) if you also want debug.

To get weights and outputs needed to run the tests dla34 and resnet101 use the Python script and the Anaconda environment included in the repository.

Create Anaconda environment and activate it:

conda env create -f file_name.yml
source activate env_name
python <script name>

To get the weights needed to run Centernet tests use this fork of the original Centernet.

git clone https://github.com/sapienzadavide/CenterNet.git

• follow the instruction in the README.md and INSTALL.md

python demo.py --input_res 512 --arch resdcn_101 ctdet --demo /path/to/image/or/folder/or/video/or/webcam --load_model ../models/ctdet_coco_resdcn101.pth --exp_wo --exp_wo_dim 512
python demo.py --input_res 512 --arch dla_34 ctdet --demo /path/to/image/or/folder/or/video/or/webcam --load_model ../models/ctdet_coco_dla_2x.pth --exp_wo --exp_wo_dim 512

To get the weights needed to run Mobilenet tests use this fork of a Pytorch implementation of SSD network.

git clone https://github.com/mive93/pytorch-ssd
cd pytorch-ssd
conda env create -f env_mobv2ssd.yml
python run_ssd_live_demo.py mb2-ssd-lite <pth-model-fil> <labels-file>

tkDNN implement and easy parser for darknet cfg files, a network can be converted with tk::dnn::darknetParser:

// example of parsing yolo4
tk::dnn::Network *net = tk::dnn::darknetParser("yolov4.cfg", "yolov4/layers", "coco.names");
net->print();

All models from darknet are now parsed directly from cfg, you still need to export the weights with the described tools in the previous section.

This is an example using yolov4.

To run the an object detection first create the .rt file by running:

rm yolo4_fp32.rt        # be sure to delete(or move) old tensorRT files
./test_yolo4            # run the yolo test (is slow)

If you get problems in the creation, try to check the error activating the debug of TensorRT in this way:

cmake .. -DDEBUG=True
make

Once you have successfully created your rt file, run the demo:

./demo yolo4_fp32.rt ../demo/yolo_test.mp4 y

In general the demo program takes 7 parameters:

./demo <network-rt-file> <path-to-video> <kind-of-network> <number-of-classes> <n-batches> <show-flag> <conf-thresh>

where

• <network-rt-file> is the rt file generated by a test
• <<path-to-video> is the path to a video file or a camera input
• <kind-of-network> is the type of network. Thee types are currently supported: y (YOLO family), c (CenterNet family) and m (MobileNet-SSD family)
• <number-of-classes>is the number of classes the network is trained on
• <n-batches> number of batches to use in inference (N.B. you should first export TKDNN_BATCHSIZE to the required n_batches and create again the rt file for the network).
• <show-flag> if set to 0 the demo will not show the visualization but save the video into result.mp4 (if n-batches ==1)
• <conf-thresh> confidence threshold for the detector. Only bounding boxes with threshold greater than conf-thresh will be displayed.

N.b. By default it is used FP32 inference

To run the an object detection demo with FP16 inference follow these steps (example with yolov3):

export TKDNN_MODE=FP16  # set the half floating point optimization
rm yolo3_fp16.rt        # be sure to delete(or move) old tensorRT files
./test_yolo3            # run the yolo test (is slow)
./demo yolo3_fp16.rt ../demo/yolo_test.mp4 y

N.b. Using FP16 inference will lead to some errors in the results (first or second decimal).

To run the an object detection demo with INT8 inference three environment variables need to be set:

• export TKDNN_MODE=INT8: set the 8-bit integer optimization
• export TKDNN_CALIB_IMG_PATH=/path/to/calibration/image_list.txt : image_list.txt has in each line the absolute path to a calibration image
• export TKDNN_CALIB_LABEL_PATH=/path/to/calibration/label_list.txt: label_list.txt has in each line the absolute path to a calibration label

You should provide image_list.txt and label_list.txt, using training images. However, if you want to quickly test the INT8 inference you can run (from this repo root folder)

bash scripts/download_validation.sh COCO

to automatically download COCO2017 validation (inside demo folder) and create those needed file. Use BDD instead of COCO to download BDD validation.

Then a complete example using yolo3 and COCO dataset would be:

export TKDNN_MODE=INT8
export TKDNN_CALIB_LABEL_PATH=../demo/COCO_val2017/all_labels.txt
export TKDNN_CALIB_IMG_PATH=../demo/COCO_val2017/all_images.txt
rm yolo3_int8.rt        # be sure to delete(or move) old tensorRT files
./test_yolo3            # run the yolo test (is slow)
./demo yolo3_int8.rt ../demo/yolo_test.mp4 y

N.B.

• Using INT8 inference will lead to some errors in the results.
• The test will be slower: this is due to the INT8 calibration, which may take some time to complete.
• INT8 calibration requires TensorRT version greater than or equal to 6.0
• Only 100 images are used to create the calibration table by default (set in the code).

export TKDNN_BATCHSIZE=2
# build tensorRT files

This will create a TensorRT file with the desired max batch size. The test will still run with a batch of 1, but the created tensorRT can manage the desired batch size.

This will test the network with random input and check if the output of each batch is the same.

./test_rtinference <network-rt-file> <number-of-batches>
# <number-of-batches> should be less or equal to the max batch size of the <network-rt-file>

# example
export TKDNN_BATCHSIZE=4           # set max batch size
rm yolo3_fp32.rt                   # be sure to delete(or move) old tensorRT files
./test_yolo3                       # build RT file
./test_rtinference yolo3_fp32.rt 4 # test with a batch size of 4

To compute mAP, precision, recall and f1score, run the map_demo.

A validation set is needed. To download COCO_val2017 (80 classes) run (form the root folder):

bash scripts/download_validation.sh COCO

To download Berkeley_val (10 classes) run (form the root folder):

bash scripts/download_validation.sh BDD

To compute the map, the following parameters are needed:

./map_demo <network rt> <network type [y|c|m]> <labels file path> <config file path>

where

• <network rt>: rt file of a chosen network on which compute the mAP.
• <network type [y|c|m]>: type of network. Right now only y(yolo), c(centernet) and m(mobilenet) are allowed
• <labels file path>: path to a text file containing all the paths of the ground-truth labels. It is important that all the labels of the ground-truth are in a folder called 'labels'. In the folder containing the folder 'labels' there should be also a folder 'images', containing all the ground-truth images having the same same as the labels. To better understand, if there is a label path/to/labels/000001.txt there should be a corresponding image path/to/images/000001.jpg.
• <config file path>: path to a yaml file with the parameters needed for the mAP computation, similar to demo/config.yaml

Example:

cd build
./map_demo dla34_cnet_FP32.rt c ../demo/COCO_val2017/all_labels.txt ../demo/config.yaml

This demo also creates a json file named net_name_COCO_res.json containing all the detections computed. The detections are in COCO format, the correct format to submit the results to CodaLab COCO detection challenge.

This branch should work on every NVIDIA GPU supported in windows with the following dependencies:

• WINDOWS 10 1803 or HIGHER
• CUDA 10.0 (Recommended CUDA 11.2 )
• CUDNN 7.6 (Recommended CUDNN 8.1.1 )
• TENSORRT 6.0.1 (Recommended TENSORRT 7.2.3.4 )
• OPENCV 3.4 (Recommended OPENCV 4.2.0 )
• MSVC 16.7
• YAML-CPP
• EIGEN3
• 7ZIP (ADD TO PATH)
• NINJA 1.10

All the above mentioned dependencies except 7ZIP can be installed using Microsoft's VCPKG . After bootstrapping VCPKG the dependencies can be built and installed using the following command :

opencv4(normal) - vcpkg.exe install opencv4[tbb,jpeg,tiff,opengl,openmp,png,ffmpeg,eigen]:x64-windows yaml-cpp:x64-windows eigen3:x64-windows --x-install-root=C:\opt --x-buildtrees-root=C:\temp_vcpkg_build

opencv4(cuda) - vcpkg.exe install opencv4[cuda,nonfree,contrib,eigen,tbb,jpeg,tiff,opengl,openmp,png,ffmpeg]:x64-windows yaml-cpp:x64-windows eigen3:x64-windows --x-install-root=C:\opt --x-buildtrees-root=C:\temp_vcpkg_build

To build opencv4 with cuda and cudnn version corresponding to your cuda version,vcpkg's cudnn portfile needs to be modified by adding $ENV{CUDA_PATH} at lines 16 and 17 in the portfile.cmake

After VCPKG finishes building and installing all the packages delete C:\temp_vcpkg_build and add C:\opt\x64-windows\bin and C:\opt\x64-windows\debug\bin to path

tkDNN is built with cmake(3.15+) on windows along with ninja.Msbuild and NMake Makefiles are drastically slower when compiling the library compared to windows

git clone https://github.com/ceccocats/tkDNN.git
cd tkdnn-windows
mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Release -G"Ninja" ..
ninja -j4

This example uses yolo4_tiny.
To run the object detection file create .rt file bu running:

.\test_yolo4tiny.exe

Once the rt file has been successfully create,run the demo using the following command:

.\demo.exe yolo4tiny_fp32.rt ..\demo\yolo_test.mp4 y 

For general info on more demo paramters,check Run the demo section on top To run the test_all_tests.sh on windows,use git bash or msys2

This is an untested feature on windows.To run the object detection demo with FP16 interference follow the below steps(example with yolo4tiny):

set TKDNN_MODE=FP16
del /f yolo4tiny_fp16.rt
.\test_yolo4tiny.exe
.\demo.exe yolo4tiny_fp16.rt ..\demo\yolo_test.mp4

To run object detection demo with INT8 (example with yolo4tiny):

set TKDNN_MODE=INT8
set TKDNN_CALIB_LABEL_PATH=..\demo\COCO_val2017\all_labels.txt
set TKDNN_CALIB_IMG_PATH=..\demo\COCO_val2017\all_images.txt
del /f  yolo4tiny_int8.rt        # be sure to delete(or move) old tensorRT files
.\test_yolo4tiny.exe           # run the yolo test (is slow)
.\demo.exe yolo4tiny_int8.rt ..\demo\yolo_test.mp4 y

Mobilenet and Centernet demos work properly only when built with msvc 16.7 in Release Mode,when built in debug mode for the mentioned networks one might encounter opencv assert errors

All Darknet models work properly with demo using MSVC version(16.7-16.9)

It is recommended to use Nvidia Driver(465+),Cuda unknown errors have been observed when using older drivers on pascal(SM 61) devices.

1. Redmon, Joseph, and Ali Farhadi. "YOLO9000: better, faster, stronger." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
2. Redmon, Joseph, and Ali Farhadi. "Yolov3: An incremental improvement." arXiv preprint arXiv:1804.02767 (2018).
3. Yu, Fisher, et al. "Deep layer aggregation." Proceedings of the IEEE conference on computer vision and pattern recognition. 2018.
4. Zhou, Xingyi, Dequan Wang, and Philipp Krähenbühl. "Objects as points." arXiv preprint arXiv:1904.07850 (2019).
5. Sandler, Mark, et al. "Mobilenetv2: Inverted residuals and linear bottlenecks." Proceedings of the IEEE conference on computer vision and pattern recognition. 2018.
6. He, Kaiming, et al. "Deep residual learning for image recognition." Proceedings of the IEEE conference on computer vision and pattern recognition. 2016.
7. Wang, Chien-Yao, et al. "CSPNet: A New Backbone that can Enhance Learning Capability of CNN." arXiv preprint arXiv:1911.11929 (2019).
8. Bochkovskiy, Alexey, Chien-Yao Wang, and Hong-Yuan Mark Liao. "YOLOv4: Optimal Speed and Accuracy of Object Detection." arXiv preprint arXiv:2004.10934 (2020).
9. Bochkovskiy, Alexey, "Yolo v4, v3 and v2 for Windows and Linux" (https://github.com/AlexeyAB/darknet)
10. Wang, Chien-Yao, Alexey Bochkovskiy, and Hong-Yuan Mark Liao. "Scaled-YOLOv4: Scaling Cross Stage Partial Network." arXiv preprint arXiv:2011.08036 (2020).