Untitled.video.-.Made.with.Clipchamp.mp4

Model Changes

Inside saved_model/1/ , there are two folders named variables/ and saved_model.pb.

Inside variables/ , there are two files, namely variables.index and variables.data-00000-of-00001.

Since github doesn't allow large files, variables.data-00000-of-00001 can't be uploaded. Download it from the last Session

We need both variables/ and model.pb to utilize the model.

Note that in saved_model/1/variables/ we need to add the model's appropriate logged variables

saved_model.pb can be replaced with any pre-trained model. Try using different models e.g SSD RESNET but must be provided with according variables and model.pb

Server Changes

Note that Changing Inference Server may affect the Model Format

Use the Model Format accordingly which is supported by the type of Inference Server

RUN MANUALLY

1. simply create new directory and pull this repo

2. To run the TF-Serving, you must have already installed docker and TF-Serving image

3. After that bind the ports: and docker run that container by this cmd:

    sudo docker run --gpus all -p 8500:8500 --name od --mount type=bind,source=/home/...USR_DIR.../tf_obj_detect_faster_rcnn/saved_model,target=/models/od -e MODEL_NAME=od -t tensorflow/serving:latest-gpu 
   

4. At last simply run 'app.py' and upload the images, TF-serving will inference and give back the results

    python app.py 
   

RUN EASY WAY (DOCKER_COMPOSE)

1. run --build to build docker compose

   sudo docker compose up --build
   

2. run docker compose up to create container images and up-running, if you have already run sudo docker compose up --build , you can neglect the below cmd:

    sudo docker compose up
   

3. stop and remove container images:

    sudo docker compose down
   

Download variables.data-00000-of-00001 here