QuailModel sample: QM Synthetic Fruit with Ripeness Labels (6 fruits)

README.md

@@ -0,0 +1,48 @@
+---
+license: cc-by-nc-sa-4.0
+task_categories:
+- object-detection
+tags:
+- fruit
+- ripeness
+- agriculture
+- food
+- synthetic
+- yolo
+pretty_name: "QM Synthetic Fruit with Ripeness Labels (6 fruits)"
+size_categories:
+- n<1K
+---
+# QM Synthetic Fruit with Ripeness Labels (6 fruits) - Free Sample
+
+Synthetic **fruit + ripeness** images: apples, oranges, lemons, bananas, tomatoes and pears with pixel-exact YOLO
+boxes, and every fruit labelled with a **ripeness stage** (unripe / ripe / overripe, driving skin colour) and
+blemish count in `meta/` - for agritech sorting lines, retail shelf monitoring, food-waste and robotics picking.
+Clean, semi-stylized Cycles renders (smooth lathe-profiled fruit; bananas are segment-built and slightly lumpy).
+Fine-tune on real photos for production use.
+
+## Contents
+| split | images | boxes |
+|---|---|---|
+| train | 247 | 1298 |
+| val | 31 | 178 |
+| test | 22 | 130 |
+
+- Image size: 640x640 JPEG. Labels: YOLO txt, classes: 0 = apple, 1 = orange, 2 = lemon, 3 = banana, 4 = tomato, 5 = pear. Images with an empty label file are true negatives (3 of 300).
+- Boxes per class: apple 259, orange 285, lemon 280, banana 205, tomato 280, pear 297
+- `meta/` holds per-object attribute JSON (class, box, visibility and module-specific attributes).
+- `data.yaml` included - train directly with Ultralytics YOLO.
+- Box size distribution (fraction of image width): median 0.117, 0% of boxes are smaller than 16 px.
+
+## How it was made
+- Procedural generator: QuailModel dataset factory, fruit module (original code; Chaikin-smoothed lathe profiles, noise-textured skins, ripeness-dependent colour, surface blemishes).
+- Labels from exact projected mesh vertices; occlusion checked by ray casting.
+- Train/val/test split 80/10/10 by file-name hash.
+
+
+## Licence
+Sample edition: CC BY-NC-SA 4.0 (non-commercial). The full commercial edition (5000 images, commercial licence) is available from QuailModel: https://huggingface.co/quailrobot
+
+## Disclosure
+Generated with AI assistance: the procedural generator code was written with an AI model and every
+image is computer-rendered. Synthetic data should be validated on real footage before deployment.

data.yaml

@@ -0,0 +1,11 @@
+path: .
+train: images/train
+val: images/val
+test: images/test
+names:
+  0: apple
+  1: orange
+  2: lemon
+  3: banana
+  4: tomato
+  5: pear