Bigthinx Object Detection Hiring Challenge
The aim of this challenge is to train an object detection model on the given dataset.
Welcome to Bigthinx Object Detection AI Challenge!
Bigthinx is hiring a full time data scientist via this AI challenge. All you need to do is to make a submission to be eligible for an interview. For this position Bigthinx is looking for an individual with an experience of 2 - 3 years.
The aim of the challenge is to train an object detection model on the given dataset. The currently provided dataset consists of 2 classes: person and car. The training annotations are given in a similar format to the COCO Object detection dataset. Please refer to "train.csv" to see the exact format values.
The training data includes images and box annotation coordinates provided in the form of CSV. More details regarding submissions can be found in the: "Submission Guidelines" section. Your model should make predictions on the "test" directory images.
NOTE: The model may only be trained on given training data, no additional data can be used for training. Although, you may freely use techniques such as Image Augmentation.
- You are required to submit your test set prediction file: output.csv - read below how to generate this file
- IPython Notebook describing your approach and the code. The notebook must be uploaded in the "My Submissions" section of this challenge.
- Upload your resume.
- Train/re-train your model on the provided dataset only. Transfer learning is allowed.
The provided dataset includes two directories "train" (contains 2,161 training images) and "test" (contains 516 test images) and two files named "train.csv" (containing training annotations) and "sample_submission.csv". "train.csv" contains 8 columns representing annotations for objects in given images in the following order:
Important: By design, not all the annotations are provided in the training set. It's normal to find only one annotation per image.
The final output should be:
Single file "output.csv" containing 9 attributes: filename (name of images), width (width of image), height (height of image), class, xmin, ymin, xmax, ymax (bounding box co-ordinates), and conf (confidence of prediction) (also referred as the probability of prediction in some cases).
The order of the attributes must be kept exactly the same, refer to "sample_submission.csv" for help.
Please note carefully: Each row MUST contain details of a single object only. Hence in the case of multiple objects in an image, they have to be inserted as each row separately with the same filename.
A sample screenshot of the "Sample Submission" is given below for reference.
How to make a submission?
- Click on "My Submission"
- On the next page, click on "+ New Submission"
- Upload your CSV in the next page and click on "Submit for Review"
- You must submit your CSV file by uploading the CSV in the "My Submissions" section of this challenge.
- Your submission will be auto-graded and you will be able to see your results instantly.
- If there is an error in the submission, your final score will be marked as 0.
- mAP, Intersection over Union (IoU)
- We are using an automatic anti-cheat system. Our system flags submission which has low trust scores. Manually modifying your submission CSV (via comparison of any form etc.), using image comparison techniques (pixel matching, file size matching, etc,), in any form, can lead to your disqualification without any notice.
- Submission must not include copyrighted code. If the violation is found, the submission will be rejected.
- The submission should be in a proper format as described by "Submission Guidelines".
- Late submission will not be accepted beyond the provided deadline (Indian Standard Time).
NOTE: We may request the code files if there is any discrepancy in your score. Your score will be marked invalid or you can be disqualified if the request for code is not fulfilled by you.
- Invitation for interview for Full-time hiring with CTC of 6 Lakhs to 10 Lakhs.
- The participants with a successful submission will get a certificate of participation.
- The top 3 participants will receive a permanent place in Dockship's "Hall of Fame".