CamelliaOfSylhet: An In-the-Wild Tea Leaf Image Dataset.

Published: 25 October 2026| Version 1 | DOI: 10.17632/7nmvdyr2g4.1
Contributors:
,
,

Description

CamelliaOfSylhet is an unlabelled, in-the-wild image dataset of tea leaves collected from tea gardens in Sylhet, Bangladesh. The images were acquired under natural field conditions, capturing variations in lighting, backgrounds, viewpoints, and leaf appearance. The dataset includes no class labels, bounding boxes, segmentation masks, or plant- and garden-level identifiers. š——š—”š—§š—”š—¦š—˜š—§ š—¦š—§š—”š—§š—œš—¦š—§š—œš—–š—¦ • Total images: 1,637 • HEIC images: 1,504 • JPEG images: 133 • Collection location: Sylhet, Bangladesh • Image environment: Natural tea-garden conditions • Annotation status: Unlabelled • Image type: Original field photographs š—œš— š—”š—šš—˜ š—”š—–š—¤š—Øš—œš—¦š—œš—§š—œš—¢š—” š——š—˜š—©š—œš—–š—˜ • Manufacturer: Samsung • Model: Galaxy M35 5G š—£š—¢š—§š—˜š—”š—§š—œš—”š—Ÿ š—„š—˜š—¦š—˜š—”š—„š—–š—› š—”š—£š—£š—Ÿš—œš—–š—”š—§š—œš—¢š—”š—¦ • š—¦š—²š—¹š—³-š˜€š˜‚š—½š—²š—æš˜ƒš—¶š˜€š—²š—± š—¹š—²š—®š—æš—»š—¶š—»š—“: Learning visual representations from unlabelled tea-leaf images. • š—Øš—»š˜€š˜‚š—½š—²š—æš˜ƒš—¶š˜€š—²š—± š—½š—æš—²š˜š—æš—®š—¶š—»š—¶š—»š—“: Exploring feature extraction and representation learning without class annotations. • š——š—¼š—ŗš—®š—¶š—»-š˜€š—½š—²š—°š—¶š—³š—¶š—° š—½š—æš—²š˜š—æš—®š—¶š—»š—¶š—»š—“: Pretraining or adapting vision models on unlabelled tea-leaf imagery to learn domain-relevant visual representations for subsequent fine-tuning on annotated datasets. • š—Ÿš—²š—®š—³ š—±š—²š˜š—²š—°š˜š—¶š—¼š—»: Developing detection models after adding suitable bounding-box annotations. • š—Ÿš—²š—®š—³ š˜€š—²š—“š—ŗš—²š—»š˜š—®š˜š—¶š—¼š—»: Creating segmentation datasets through additional pixel-level annotation. • š—©š—¶š˜€š˜‚š—®š—¹ š—³š—²š—®š˜š˜‚š—æš—² š—®š—»š—®š—¹š˜†š˜€š—¶š˜€: Investigating variations in leaf appearance under natural field conditions. • š——š—®š˜š—®š˜€š—²š˜ š—±š—²š˜ƒš—²š—¹š—¼š—½š—ŗš—²š—»š˜: Supporting the creation of expert-annotated tea-leaf datasets for future agricultural computer-vision research. These applications are potential research directions and have not been experimentally evaluated as part of this dataset release. š—žš—˜š—¬š—Ŗš—¢š—„š——š—¦ CamelliaOfSylhet; Tea Leaf Dataset; Camellia sinensis; Sylhet; Bangladesh; In-the-Wild Images; Unlabelled Dataset; Agricultural Computer Vision; Self-Supervised Learning; Unsupervised Pretraining; Domain-Specific Pretraining; Leaf Detection; Leaf Segmentation.

Files

Institutions

Categories

Computer Vision, Tea, Deep Learning, Transfer Learning, Self-Supervised Learning, Agriculture

Licence