README.md

metadata

task_categories:
  - image-segmentation
  - image-classification
language:
  - en
tags:
  - agritech
  - hyperspectral
  - spectroscopy
  - fruit
  - sub-class classification
  - detection
size_categories:
  - 10K<n<100K
license: mit

Living Optics Hyperspectral Fruit Dataset

Overview

This dataset contains 100 images of various fruits and vegetables captured under controlled lighting, with the Living Optics Camera.

The data consists of RGB images, sparse spectral samples and instance segmentation masks.

From the 100 images, we extract >430,000 spectral samples, of which >85,000 belong to one of the 19 classes in the dataset. The rest of the spectra can be used for negative sampling when training classifiers.

An additional 11 labelled images are provided as a validation set.

Additionally, we provide a set of demo videos in .lo format which are unannotated but which can be used to qualititively test algorithms built on this dataset.

Classes

The training dataset contains 19 classes:

• 🍋 lemon - 8275 total spectral samples
• 🍈 melon - 9507 total spectral samples
• 🥒 cucumber - 227 total spectral samples
• 🍏 granny smith apple - 3984 total spectral samples
• 🍏 jazz apple - 272 total spectral samples
• 🍎 plastic apple - 6693 total spectral samples
• 🍎 pink lady apple - 17311 total spectral samples
• 🍎 royal gala apple - 21319 total spectral samples
• 🍅 tomato - 3748 total spectral samples
• 🍅 cherry tomato - 360 total spectral samples
• 🍅 plastic tomato - 569 total spectral samples
• 🫑 green pepper - 226 total spectral samples
• 🫑 yellow pepper - 4752 total spectral samples
• 🫑 orange pepper - 552 total spectral samples
• 🍊 orange - 4641 total spectral samples
• 🍊 easy peeler orange - 2720 total spectral samples
• 🍐 pear - 194 total samples
• 🍇 green grape - 106 total spectral samples
• 🍋‍🟩 lime - 43 total spectral samples

Requirements

• lo-sdk
• lo-data

Download instructions

Command line

mkdir -p hyperspectral-fruit
huggingface-cli download LivingOptics/hyperspectral-fruit --repo-type dataset --local-dir hyperspectral-fruit

Python

from huggingface_hub import snapshot_download
dataset_path = snapshot_download(repo_id="LivingOptics/hyperspectral-fruit", repo_type="dataset")
print(dataset_path)

Usage

import os.path as op
import numpy.typing as npt
from typing import List, Dict, Generator
from lo.data.tools import Annotation, LODataItem, LOJSONDataset, draw_annotations
from lo.data.dataset_visualisation import get_object_spectra, plot_labelled_spectra
from lo.sdk.api.acquisition.io.open import open as lo_open

# Load the dataset
path_to_download = op.expanduser("~/Downloads/hyperspectral-fruit")
dataset = LOJSONDataset(path_to_download)

# Get the training data as an iterator 
training_data: List[LODataItem] = dataset.load("train")

# Inspect the data
lo_data_item: LODataItem
for lo_data_item in training_data[:3]:

    draw_annotations(lo_data_item)

    ann: Annotation
    for ann in lo_data_item.annotations:
        print(ann.class_name, ann.category, ann.subcategories)

# Plot the spectra for each class
fig, ax = plt.subplots(1)
object_spectra_dict = {}
class_numbers_to_labels = {0: "background_class"}
for lo_data_item in training_data:
    object_spectra_dict, class_numbers_to_labels = get_object_spectra(
        lo_data_item, object_spectra_dict, class_numbers_to_labels
    )

plot_labelled_spectra(object_spectra_dict, class_numbers_to_labels, ax)
plt.show()    

See our Spatial Spectral ML project for an example of how to train and run a segmentation and spectral classification algoirthm using this dataset.