https://github.com/varun0157/semantic-segmentation
Semantic Segmentation using FCN and UNet models [Assignment 4 of Computer Vision, IIIT-Hyderabad, Spring '25].
https://github.com/varun0157/semantic-segmentation
fcn semantic-segmentation unet
Last synced: 9 months ago
JSON representation
Semantic Segmentation using FCN and UNet models [Assignment 4 of Computer Vision, IIIT-Hyderabad, Spring '25].
- Host: GitHub
- URL: https://github.com/varun0157/semantic-segmentation
- Owner: Varun0157
- Created: 2025-04-28T17:07:49.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-10-03T13:32:31.000Z (9 months ago)
- Last Synced: 2025-10-03T15:29:44.275Z (9 months ago)
- Topics: fcn, semantic-segmentation, unet
- Language: Python
- Homepage:
- Size: 22.1 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
Semantic Segmentation using FCN and UNet models, done as a part of Assignment 4 of the Computer Vision course (IIIT-Hyderabad, Spring '25). The assignment details can be found in the [assignment document](./docs/CV_S25_A4.pdf).
## Results
### FCN
For ablations and visualisations, see [the report](./FCN/report.pdf).
#### Visualisations
_Each image has three columns, representing input image, ground truth, followed by prediction (each image is a collage of these three in that order)._
| Backbone type | 32s | 16s | 8s |
| ------------- | --------------------------------------------------------- | --------------------------------------------------------- | -------------------------------------------------------- |
| Frozen | 
| 
| 
|
| Unfrozen | 
| 
| 
|
To train and test the model, run:
```bash
cd FCN
python -m src.main
```
The options include the data directory, batch size, model variant, frozen vs unfrozen backbone, the number of epochs, and the learning rate. See the [`main.py`](./FCN/src/main.py) file for more details.
To visualise some test results on a checkpoint, run:
```bash
cd FCN
./visualise.sh
```
### UNet
For ablations and visualisations, see [the report](./UNet/report.pdf).
#### Visualisations
_Each image has three columns, representing input image, ground truth, followed by prediction (each image is a collage of these three in that order)._
| Variant | Prediction 1 | Prediction 2 |
| --------------- | ------------------------------------------------------- | ------------------------------------------------------- |
| Vanilla | 
| 
|
| Residual | 
| 
|
| No-Skip | 
| 
|
| Gated Attention | 
| 
|
To train and test the model, run:
```bash
cd UNet
python -m src.main
```
The options include the data directory, batch size, model variant, the number of epochs, and the learning rate. See the [`main.py`](./UNet/src/main.py) file for more details.
To visualise some test results on a checkpoint, run:
```bash
cd UNet
./visualise.sh
```
## Setup
Download the data from [the following link](https://drive.google.com/drive/folders/1s2ZgwawnZyZXc5eei5cWmgV2A7UXOrMV?usp=sharing).
The environment can be set up with the conda env file:
```bash
cd docs
conda env create -f env.yml
```
or
```bash
pip install -r requirements.txt
```
Alternatively, install the dependencies as in [the conda history](./docs/env-hist.yml).
### Scripts
Helper utilities are available to split the train data [into train and validation](./FCN/scripts/split.py) for the FCN, and [for the UNet](./UNet/scripts/split.py), as well as to perform some [exploratory data analysis](./FCN/scripts/eda.py) for the FCN.