jetson-interface retrain

Autonomous Machines Jetson & Embedded Systems Jetson TX2
1

I am just wondering if that is possible to retrain camera-video-detection without DIGITS, but naively on Jetson.

e.g. with tensorflow I do use Jetson [I do not have a pc to run DIGITS at] for static images recognition that way:

I
sudo bazel build tensorflow/examples/image_retraining:retrain

II
sudo bazel-bin/tensorflow/examples/image_retraining/retrain --image_dir=/home/nvidia/Downloads/images/

III
sudo bazel build tensorflow/examples/label_image:label_image && bazel-bin/tensorflow/examples/label_image/label_image --graph=/tmp/output_graph.pb --labels=/tmp/output_labels.txt --output_layer=final_result --image=/home/nvidia/Downloads/Test/TS_1384_0.jpg --input_layer=Mul

References:
https://github.com/dusty-nv/jetson-inference#running-the-live-camera-detection-demo-on-jetson
https://github.com/dusty-nv/jetson-inference#re-training-the-network-with-digits
https://www.tensorflow.org/tutorials/image_retraining

2

Maybe my previous question is helpful?

[url]https://devtalk.nvidia.com/default/topic/1022718/jetson-tx2/create-object-detection-model-without-digits-/[/url]

3

thank you for sharing the link

4

Do you mean that if I compile the nvcaffe - there is a way to find a simple way to retrain live-detection of objects?
I shall look into that.

5

perhaps the resource Caffe Deep Learning Framework - NVIDIA Jetson TX2 - JetsonHacks can be used as a reference for installation

6 
git clone https://github.com/NVIDIA/caffe

cd caffe

git checkout caffe-0.15

editing the file [edit file $CAFFE_ROOT/3rdparty/cub/host/mutex.cuh]

line124: --- #ifndef __arm__
line124: +++ #if !defined(__arm__) && !defined(__aarch64__)

[AAstaLLL]

now an excerpt from jetsonhacks

sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev \
libhdf5-serial-dev protobuf-compiler -y

sudo apt-get install --no-install-recommends libboost-all-dev -y

sudo apt-get install libatlas-base-dev -y

sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev -y

sudo apt-get install python-dev python-numpy -y

cp Makefile.config.example Makefile.config

cmake -DCUDA_USE_STATIC_CUDA_RUNTIME=OFF

make -j6 all

References:
https://devtalk.nvidia.com/default/topic/1022718/jetson-tx2/create-object-detection-model-without-digits-/post/5204190/#5204190
https://devtalk.nvidia.com/default/topic/976063/jetson-tx1/the-problem-with-the-assembly-of-caffe-0-15-on-jetsontx1/post/5019508/#5019508

7

Hi,

Please use NvCaffe to train DetectNet.
Installation guide is here:
[url]https://devtalk.nvidia.com/default/topic/1021931/jetson-tx2/nvcaffe-support-on-tx2/post/5200686/#5200686[/url]