Coding Period: Week 2

1 minute read


So basically, the first week of the coding period ididn’t go as well as planned, and I realised that you need to give MUCHH MORE TIMEE to the work that you are doing, and hence, I decided to work on a remaining task of week 1 for week 2 as well. Now, as per the week 2 tasks, I feel that the TensorFlow Object detection one should be easier, as I have TRIED (read Tried, not Worked xD) it before, and it gave me a hugeee headache. But now, i’ll be more focused and work on it more. As per the reading stuff part, that should be easy, but time-consuming, as the only thing that I do is search my errors on StackOverflow.
Let’s see how week 2 goes!


  • Read something about Models that are specifically used for human detection
  • [] Understand and work on TFOD (TensorFlow Object Detection), and if time permits, learn PyTorch (like atleast be in a position to understand code in PyTorch)
  • Convert a model from TensorFlow(Models from TFOD) to ONNX
  • Check the compatibility of both formats for Robotics Academy

The execution

My thought process was to start reading about the Models that are Specifically used for Human Detetction. Well, across this, I came about the Tiny YOLO model, the SSD_Volumenet model as well as different CNN models such as R_CNN, Faster R_CNN and ShuffleNet as well.

After reading about the different models that are used specifically for Human Detection, I decided to install the TensorFlow Object Detection API. So, I first cloned the Git repository [6] to my directory, and then installed the Protobuffers and the COCO API. This link [7] helped out a lot for this purpose.

The next work was for me to convert a model from the TFOD API to ONNX, and since the models that were picked up for the Human Detection exercise were available in the ONNX Modle Zoo, I didn’t do this. It was collectively decided to use the ONNX Model format for the Exercise.


[6] [] (
[7] [] (\