Helmet Detection AI Challenge
Challenge requires to develop high accuracy object detection on helmets.(Major Focus : Accuracy)
Interns
This task requires you to develop high accuracy model for helmet detection. The techniques used will be judged on the basis of accuracy.
The candidates will be invited for an Internship Interview based on their performance on the leader-board.
NOTE:
The dataset provided is a collection of sample helmet images from the internet without annotations, so the candidate will have to use a separate dataset for training.
The submission should include:
- Jupyter Notebook (.ipynb) showing the approach taken for training and the dataset used.
- Summary Report (approx. 500-1000 words) explaining the challenges faced during the challenge and the approach taken to solve them.
- 'requirements.txt' with list of required packages.
- Execution Script for detection for test images (.py) (run.py) file.
The execution script should be named “run.py” and will be supplied a “--testdir” named command line argument for evaluation, which will contain the images in "jpg" format, the example is given below:
run.py —testdir <path to test directory>
The results must be in JSON Format (output.json) with location of helmets in an image using top left and bottom right co-ordinate.
Example:
{
"Name of Image": "[
[helmet 1 top-left co-ordinate, helmet 1 bottom right],
[helmet 2 top-left co-ordinate, helmet 2 bottom right],
...
]",
...
}
{
"1.jpg": "[
[(10, 100), (40, 200)],
[(50, 200), (90, 3500)],
...
]",
...
}
Currently only Python-framework based models are accepted:
- Tensorflow
- Pytorch
- Keras
Judgement
- Detection Accuracy
Rules
- Submission must not include copyrighted code. If violation is found, submission will be rejected.
- The submission should be in a proper format as described by "Submission Guidelines".
- Late submission will not be accepted beyond provided deadline (Indian Standard Time).
The candidates will be invited for an Internship Interview based on their performance.
Certificates will be provided after successful submission of a solution to this challenge from dockship.io
Challenge Started
Challenge Ended