pierluigiferrari/ssd_keras

Health v2 · maintenance only

Adoption not part of the score

1869 stars · 918 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

A Keras implementation of the Single Shot MultiBox Detector (SSD) object detection architecture, with ports of the original trained weights for SSD300, SSD512, and a small SSD7 variant. It is heavily documented and intended for users who want to understand, adapt, or fine-tune the model at a low level.

Use cases

• train an SSD object detection model in Keras
• fine-tune pretrained SSD weights on my own dataset
• run object detection with a pretrained SSD300 or SSD512 model
• learn how the SSD architecture works at a low level
• build a custom SSD with my own base network
• train a small fast object detector on a mid-tier GPU

When to choose

• you want an accurate, well-documented SSD implementation in Keras with matching mAP to the original Caffe models
• you need pretrained SSD weights for transfer learning or fine-tuning
• you want to study or modify the SSD architecture in detail

When to avoid

• you need state-of-the-art detectors like YOLO or transformers instead of SSD
• you use PyTorch rather than Keras/TensorFlow
• you need actively maintained code with support for the latest TensorFlow versions

Facets

library · maturity maintenance

machine-learning deep-learning computer-vision image-processing computer-vision deep-learning machine-learning image-processing python cross-platform ssd object-detection keras tensorflow single-shot-multibox-detector vgg16 transfer-learning pascal-voc gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="pierluigiferrari/ssd_keras")