tadejmagajna / hereiswally Goto Github PK

There's missing file for the model retraining, following the guide the file missing is The pipeline config.
The model is not accurate as it can't recognize waldo if is bigger than a threshold, the images where waldo is big the model say is not found.

Hi there! Great work! I was curious about how to print the x y location of Waldo in your Python script, if you can give me some hints will be really appreciate!

In the first section "Preparing the dataset", you mention the labels corresponding to the images that would be packed into a single binary TF file.

I speculate this labels file is a collection of annotations in PASCAL VOC format, output from a program like labelImg.

This file seems to be missing from the repo.

Looking in final.config, is's possible that one of the two following files is our missing labels file.

label_map_path: "/Users/tadej/waldo/labels.pbtxt"
label_map_path: "/Users/tadej/waldo/labels.txt"

Please advise.

hi
nice article, i am trying to follow your steps but I got stuck at this point :

the trained_model folder should have not a checkpoint.ckpt file ??
currently in your repo and also in the link for the trained_model there is only a checkpoint file (without ckpt extension)

How should I set the "fine_tune_checkpoint:" value on my example.config file?

my example.config currently looks like this:

fine_tune_checkpoint: "/home/antoine/Documents/ia/HereIsWally/trained_model/checkpoint"

The utils module (likely from the TF Object Detection repo) is missing from this repo.

 HereIsWally git:(master) python3 find_wally.py 
Traceback (most recent call last):
  File "find_wally.py", line 8, in <module>
    from utils import label_map_util
ModuleNotFoundError: No module named 'utils'