CamelliaOfSylhet: An In-the-Wild Tea Leaf Image Dataset.
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. ššš§šš¦šš§ š¦š§šš§šš¦š§ššš¦ ⢠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 šš ššš ššš¤šØšš¦šš§šš¢š” ššš©ššš ā¢ Manufacturer: Samsung ⢠Model: Galaxy M35 5G š£š¢š§šš”š§ššš š„šš¦ššš„šš šš£š£ššššš§šš¢š”š¦ ā¢ š¦š²š¹š³-ššš½š²šæšš¶šš²š± š¹š²š®šæš»š¶š»š“: Learning visual representations from unlabelled tea-leaf images. ā¢ šØš»ššš½š²šæšš¶šš²š± š½šæš²ššæš®š¶š»š¶š»š“: Exploring feature extraction and representation learning without class annotations. ā¢ šš¼šŗš®š¶š»-šš½š²š°š¶š³š¶š° š½šæš²ššæš®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. ā¢ šš²š®š³ š±š²šš²š°šš¶š¼š»: Developing detection models after adding suitable bounding-box annotations. ā¢ šš²š®š³ šš²š“šŗš²š»šš®šš¶š¼š»: Creating segmentation datasets through additional pixel-level annotation. ā¢ š©š¶ššš®š¹ š³š²š®šššæš² š®š»š®š¹ššš¶š: Investigating variations in leaf appearance under natural field conditions. ā¢ šš®šš®šš²š š±š²šš²š¹š¼š½šŗš²š»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. ššš¬šŖš¢š„šš¦ 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
- Leading UniversitySylhet Division, Sylhet