Experiment 02 / Applied computer vision
Underwater image enhancement
A web application that restores clarity and color in underwater photographs using a super-resolution model.
2025
Personal project. Model integration, Flask application and interface.
01 / The problem
What it set
out to do.
Water absorbs light unevenly, so underwater photographs lose contrast and drift toward green and blue. The correction is tedious by hand and needs to happen before anyone can read detail in the image.
- 01Image upload
- 02Preprocessing
- 03Enhancement model
- 04Result & download
A simplified view of how the code is organized, not a production infrastructure diagram.
02 / The approach
Decisions beneath
the surface.
- 01
Put a trained enhancement model behind a single upload-and-wait interaction.
- 02
Handle the image work server-side so the browser only carries the picture and the result.
- 03
Keep the before and after together, so the change is something you judge rather than take on trust.
03 / Inside the repository
Where to look
in the code.
new/a.pyThe current Flask application: upload handling, enhancement and result delivery.
new/templates/Landing, upload and results screens for the enhancement flow.
model_enhance.pyThe enhancement step, wrapping the model call.
app.pyThe earlier version of the application, kept in the repository alongside the rewrite.
input1–4.pngSample underwater images for trying the pipeline without your own photographs.
04 / Technologies
The working set.
05 / What it left behind
A working demonstration of a deep-learning model doing one job end to end, from upload to downloadable result.
Source notes
Described from the repository's README and source tree. No accuracy benchmark or published evaluation is claimed.