Update README.md

README.md

@@ -38,64 +38,34 @@ dataset_info:
   dataset_size: 12169999
 ---
 
-# Yannik019/llm_pack_detection
+# BagBuddy HF Dataset Export
 
-This repository contains the BagBuddy dataset as a raw folder upload.
+## Files
 
-## Layout
+- `train.parquet`: self-contained parquet export with embedded image bytes
+- `dataset_info.json`: metadata produced by the `datasets` library
+- `saved_dataset/`: optional `save_to_disk` export for local reloading
 
-The root directory is organized into bucket folders such as `6`, `8`, `10`, and so on.
-Each bucket contains timestamped capture folders. Every capture folder is expected to
-contain:
+## Schema
 
-- `debug.jpg`
-- `view.jpg`
-- `depth.npy`
-- `groceries.json`
+- `image`: the `view.jpg` scene image
+- `caption`: generated caption summarizing the scene labels
+- `bucket`: top-level source bucket directory such as `6` or `10`
+- `sample_id`: timestamped sample folder name
+- `labels`: grocery item labels from `groceries.json`
+- `annotation_indices`, `annotation_x`, `annotation_y`: annotation metadata
+- `annotation_count`: number of annotated items in the scene
 
-The `groceries.json` file stores a list of labeled grocery items with at least the
-fields `name`, `idx`, `x`, and `y`.
+## Hub Usage
 
-## Summary
+```python
+from datasets import load_dataset
 
-- Samples: 40
-- Total annotations: 520
-- Buckets: 8
-- Unique labels: 18
+dataset = load_dataset("Yannik019/llm_pack_detection", split="train")
+print(dataset)
+```
 
-### Samples Per Bucket
 
-- 10: 5 samples
-- 12: 5 samples
-- 14: 5 samples
-- 16: 5 samples
-- 18: 5 samples
-- 20: 5 samples
-- 6: 5 samples
-- 8: 5 samples
+## Row count
+- Rows: 40
 
-### Label Frequencies
-
-- Joghurt Cup: 55
-- Energy Drink Can: 36
-- Noodles in Plastic Bag: 35
-- Bananas: 35
-- Strawberries: 33
-- Glass Beer Bottle: 32
-- Eggs: 32
-- Tomatoes: 31
-- Apples: 29
-- Onions: 29
-- 1.5L Plastic Water Bottle: 29
-- Mushrooms: 28
-- Canned Corn: 26
-- Lettuce: 22
-- Cucumber: 21
-- Bell Pepper: 19
-- Canned Beans: 18
-- Canned Peas: 10
-
-## Upload Notes
-
-This dataset was uploaded with `upload_dataset_to_huggingface.py`. Update this card
-if you want to document licensing, splits, or benchmark details more precisely.