DnaRnaProteins/biomech-feedback

---
configs:
- config_name: default
data_files:
- split: train
path: data/**/*.jsonl
dataset_info:
features:
- name: id
dtype: string
- name: timestamp
dtype: string
- name: model_version
dtype: string
- name: source_dataset
dtype: string
- name: image_b64
dtype: string
- name: caption
dtype: string
- name: detections
list:
- name: label
dtype: string
- name: bbox
sequence: int64
- name: segmentation_quality
dtype: int64
- name: detection_quality
dtype: int64
- name: caption_quality
dtype: int64
- name: overall_verdict
dtype: string
- name: caption_correction
dtype: string
---

# BioMech Feedback

Human feedback records for the [biomech-inference-serving](https://github.com/czi-ai/biomech-inference-serving-feedback) pipeline.

Each record captures one inference run (segmentation, detection, captioning) along with human quality ratings.

## Schema

| Field | Type | Description |
|---|---|---|
| id | string | UUID for the feedback record |
| timestamp | string | ISO-8601 UTC timestamp |
| model_version | string | Semver of the deployed model |
| source_dataset | string | Name of the source image dataset |
| image_b64 | string | Base64-encoded input image (RGB) |
| caption | string | Model-generated biomedical caption |
| detections | list of structs | Detected cells: + [x,y,w,h] |
| segmentation_quality | int64 | Human rating 1-5 (null if not rated) |
| detection_quality | int64 | Human rating 1-5 (null if not rated) |
| caption_quality | int64 | Human rating 1-5 (null if not rated) |
| overall_verdict | string | or (null if not rated) |
| caption_correction | string | Corrected caption text (null if none) |