Mounted NVIDIA docker to run a model inference with exported frozen inference graph on GPU.
151
This is a docker image to run a model inference with an exported frozen graph.
The image was used in the project "Mitigating wildlife and human conflicts in Tanzania with AI. The project is known as "AI-assisted Aerial Images Analysis (AIAIA)".
Once you run:
docker run --runtime=nvidia --gpus=all -u 0 --rm -v ${PWD}://mnt/data -it devseeddeploy/od_frozen_graph_inference:v1-tf1.15-gpu bash
Inside the docker container, compile all the images you want to run inference for, using the provided script frozen_pred.py to run:
python3 frozen_pred.py \
--frozen_inference_graph=models/livestock/frozen_inference_graph.pb \
--images_path=chips/*.jpg \
--threshold=0.5 \
--batch_size=100
The frozen_pred.py can be found at Development Seed's GitHub repo: https://github.com/developmentseed/ai4earth-wildlife-conservation/tree/master/inference
Content type
Image
Digest
Size
2.6 GB
Last updated
over 5 years ago
docker pull devseeddeploy/od_frozen_graph_inference:v1-tf1.15-gpu