Contains assignment submissions made in the course CS771 (2022-23 Sem. I) in a team of 5:
| Name | User ID |
|---|---|
| Kunwar Preet Singh | Enkryp |
| Parinay Chauhan | parinayc20 |
| Pratyush Gupta | PratyushGupta0 |
| Akhil Agrawal | akhilagrawal1001 |
| Aditya Tanwar | cliche-niche |
In each assignment, we were given a problem statement after being taught some methods/models:
- Assignment 1: Exercise on using SVMs. We explored SGDM to solve it, but ended up submitting SCDM with a bunch of case-specific optimizations.
- Assignment 2: The problem was that of multilabel classification, with complete freedom to use whatever one wants. Our group ended up using a mix of Decision Trees, and Neural Networks.
- Assignment 3: The problem required the usage of Computer Vision followed by Deep Learning, specifically CNNs, to infer greek alphabets written in a CAPTCHA. A perfectly arbitrary captcha has been provided here.
Note: We were provided with 2000 sample images for A3 but to save space on this repository, a single image has been provided.
- A1: 66/60 - 7 marks bonus for using reduced dimensionality (
<200) and 1 mark deducted fort=0.2sand0.01<=e=0.04<1(full marks fore<0.01) - A2: 87/100 - 6 marks deducted for a size of
~7.5 MB(full marks for<128 KB), 6 marks deducted for an inference time of~2s(ceil( max( 1 - t, 0.4 ) * 10 )), 1 mark deducted formprec@1 = 0.625(ceil( max( p, 0.4 ) * 5 )) - A3: 89/100 - 6 marks deducted for a size of
~11.6 MB(full marks for<128 KB), 4 marks deducted for an inference time of~30s(full marks for<5s), 1 mark deducted for inaccuracy (floor( c * 40 ))