How julilc/computer_vision works
Project type and purpose
• This is a machine-learning inference and vector-search application with a Gradio-based UI.
• It uses YOLOv8 for object detection on dog images, produces embeddings via three model variants (base, custom, fine-tuned), stores them in FAISS vector indexes, and exposes an interactive Gradio front end.
File structure
• Top-level:
• app_gradio/ – four Gradio application scripts for experimentation or different UI flows.
• models/ – serialized model artifacts: YOLOv8 (“yolov8n.pt”) plus two Keras-based custom models and two fine-tuned variants (.h5 & .keras).
• test_img/ – a YOLOv8 dataset (images + labels + YAML) for validation.
• vector_db/ – FAISS indexes and pickled vector stores for each model variant.
• dogs.csv & informe_jupyter.ipynb – auxiliary data and analysis notebook.
• No external microservices, but key “components” appear as separate directories.
README
• Empty/placeholder. No architectural guidance here, so we infer from code structure.