Pascal VOC 2012 Object Detection

Welcome to PASCAL VOC Object Detection AI Challenge

You are required to build an Deep Learning Model which has the capability of performing the task of Object Detection. You are required to build a complete training and Evaluation pipeline of Object Detection Network on the Pascal VOC 2012 Dataset.

NOTE: The challenge has been updated to support Auto-Grading. Please check "Submission Guidelines" carefully for more details.

The dataset contains 20 classes:

• airplane
• bicycle
• bird
• boat
• bottle
• bus
• car
• cat
• chair
• cow
• diningtable
• dog
• horse
• motorbike
• person
• pottedplant
• sheep
• sofa
• train
• tvmonitor

The complete details regarding the dataset can be found at:

http://host.robots.ox.ac.uk/pascal/VOC/

http://host.robots.ox.ac.uk/pascal/VOC/voc2012/index.html#devkit

Citation

The PASCALVisual Object Classes (VOC) Challenge, http://host.robots.ox.ac.uk/pascal/VOC/pubs/everingham10.pdf

The dataset is from PASCAL VOC Challenge 2012. The goal is to predict the bounding box and label of each individual object for test images provided in "pascal_test" directory. The training images are given in "pascal_train" and training labels in "labels_pascal_train.csv". There is another file "partial.csv" which contains all the fields required for the submission and pre-filled values for filenames, width and height to help avoid any confusion. The same file may be submitted as a solution after filling the predictions of other fields by the AI model.

The dataset contains 20 classes:

• airplane
• bicycle
• bird
• boat
• bottle
• bus
• car
• cat
• chair
• cow
• diningtable
• dog
• horse
• motorbike
• person
• pottedplant
• sheep
• sofa
• train
• tvmonitor

The final output should be:

Single file "output.csv" containing name of images ("filename"), width of image ("width"), height of image ("height"), "class", bounding box co-ordinates ("xmin", "ymin", "xmax", "ymax") and confidence of prediction ("conf"). The order of the attributes must be kept exactly the same, refer to partial.csv for help.

A sample output screenshot is also given below and the dataset zip also contains a "partial.csv" file which have pre-filled filenames, height and width required for the final submission.

Please note:

1. You must submit your CSV file by uploading the CSV in the "My Submissions" section of this challenge.
2. Your submission will be auto graded and you will be able to see your results instantly.
3. If there is any error in the submission, your final score will be marked as 0.

Judgement

1. mAP, Intersection over Union (IoU)

Rules

1. We are using automatic anti-cheat system. Our system flags submission which have 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.
2. Submission must not include copyrighted code. If violation is found, submission will be rejected.
3. The submission should be in a proper format as described by "Submission Guidelines".
4. Late submission will not be accepted beyond provided deadline (Indian Standard Time).

NOTE: We may request for 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.

The Top 3 participants will receive goodies from Dockship.

The goodies will include:

• T-shirt
• Premium Laptop Bag
• Personalized Mugs
• Personalized Notepads
• Laptop Stickers