catwlk

CLIP‑powered fashion engine for images of clothes.

Upload a photo and catwlk returns concise labels like type, era, occasion with confidence scores and ranked alternatives.

---

Quick start (Docker)

docker compose up --build

• Gateway (HTTP): http://localhost:8080
• CLIP RPC (TCP): localhost:9000

---

API

POST /classify

multipart/form-data

Form fields:

• image — the image file to classify

Example:

curl -F image=@look.jpg http://localhost:6060/classify

Sample response (truncated):

{
  "jsonrpc": "2.0",
  "result": {
    "type": {
      "label": "mini dress",
      "score": 0.83,
      "alternatives": [
        {"label":"mini dress","score":0.83},
        {"label":"midi dress","score":0.10}
      ]
    },
    "era": {"label":"y2k","score":0.41,"alternatives":[...]},
    "occasion": {"label":"casual","score":0.62,"alternatives":[...]}
  },
  "id": 1
}

---

Configuration (env)

Gateway (Go):

• CLIP_SERVICE_ADDRESS — TCP address of the CLIP server
  Default: clip-service:9000
• PORT — HTTP port for the gateway
  Default: 8080

The defaults work with docker-compose.yaml out of the box.

---

Local development (no Docker)

Open two terminals:

1) Start CLIP backend

cd clip-service
python3 main.py

2) Start Go gateway

cd image-to-label-service
# if your CLIP server runs elsewhere:
export CLIP_SERVICE_ADDRESS=localhost:9000
go run main.go

---

Project layout

clip-service/               # Python TCP JSON-RPC (CLIP)
  └── main.py
image-to-label-service/     # Go HTTP gateway
  └── main.go
docker-compose.yaml
public/index.html           # optional tiny demo UI

---