ML-01 Add Fish Species Classification Model

[jpdevhub]

Currently, the system assumes all scanned fish are "Rohu Carp" (hardcoded). To make the application truly versatile in real-world markets, we need to train and integrate a classification model that identifies the fish species before grading its freshness.

Tasks:

Collect an annotated dataset of common local fish species.
Train a lightweight classification model (e.g., MobileNetV3 or ResNet).
Integrate the model into the FastAPI pipeline to dynamically populate the species_detected field.

[harshadashinde2224-gif]

hi @jpdevhub
Hi, I would like to work on this issue under SSoC. Can you please assign it to me?

[jpdevhub]

Hello! Can you share your proposed approach for this issue? Specifically, details on the dataset, the classification model you intend to use and any relevant specifications before we proceed.

[SaumyaSngh323]

Hi i would like to work on this issue will update the PR in 5-6 days

[jpdevhub]

Hey @SaumyaSngh323 do one issue at a time. I have assigned you one issue do that first. For this issue love to hear the approach (dataset, model) and your experience to train models.

[arcgod-design]

Hi! I'd like to work on this issue as part of SSoC 2026.

Proposed approach:

• Dataset: I will use the "Fish-Market" or "Fish-Species" dataset from Kaggle, supplemented with web-scraped images for specific local species.
• Model: I intend to use MobileNetV3-Small due to its exceptional efficiency and low latency for Edge AI/FastAPI integration.
• Experience: I have experience training CNNs for image classification and deploying them via FastAPI using ONNX/PyTorch.

Could you please assign this issue to me? Thank you!

[jpdevhub]

@arcgod-design go for it