Deepstream for Jetson VS Deepstream for Tesla

Hi, guys

I found there are two different kinds of deepstream. One is for Jetson and another one is for tesla gpu.
There are some functions only appear in Jetson version. For example, Detect the color, model and make of the car.

What should I do to implement these function in the tesla version deepstream.

More details:
The model for these function include 5 files:
carcolorPruned.caffemodel
deploy.prototxt
labels.txt
mean.ppm
carcolorPruned.caffemodel_b2_fp16.cache

what is the 5th one, normally we only need the first 4 files for detection at most platforms, like CAFFE and YOLO.

On the deepstream for Tesla, we also only need the frist 4 files for detection.
So how can I apply these models on the deepstream for tesla?

Anybody know the answers? I am very appreciated for your help.

Best,
Yang

Ydeng,
Looks you are using Legacy deepstream for telsa, right?
We will release new deepstream for telsa soon.

carcolorPruned.caffemodel_b2_fp16.cache, this is the cache file. If we have cache file, we will use cache file directly, otherwise we will convert your model into cache file, and save it here.

If batchsize is same, it can be shared with new Deepstream for telsa.

Thanks
wayne zhu

Thanks a lot. Hope to have the new version release soon.