AgML

Dataset · Image segmentation

weed_segmentation_france

AgML

This dataset comprises real RGB images of weeds in agricultural fields at Montoldre, France, specifically within maize and bean crop areas. Captured using a robot platform during 2019 and 2021, the images were collected under natural field conditions to support semantic segmentation research for weed detection in crop environments. The dataset contains 985 images with pixel-level mask annotations. This dataset is indexed on https://project-agml.github.io/ as part of the AgML python library. This dataset was reformatted from its original format to match HuggingFace standards.

Dataset · Image classification

tree_species_classification_slovak_normal_cropped

AgML

This dataset provides real RGB images of tree species collected in natural field environments across Slovakia and the Czech Republic. Images were captured using handheld cameras (Sony Alpha 7 and Canon EOS 4000D) during field surveys in June and August 2022. The dataset contains 1,367 images across 4 classes: European beech, European silver fir, Norway spruce, Sessile oak. This dataset is indexed on https://project-agml.github.io/ as part of the AgML python library. This dataset was reformatted from its original format to match HuggingFace standards.

Dataset · Object detection

CCOD_cotton_flower_detection

AgML

CCOD (Crop Canopy Organ-level Object Detection) is a UAV dataset of small plant organs seen from above crop canopies: cotton flowers and cotton bolls in Xinjiang, potato flowers at Keshan Farm in Heilongjiang, and rice panicles in Jiangsu, China. Images were captured at flowering or maturity from 12 m with the RGB camera of a DJI Mavic 3 Multispectral drone, cropped into 640 x 640 tiles (512 x 512 for rice), and annotated with dense bounding boxes, using a SAHI-YOLOv10n model to propose boxes that annotators then corrected. The cropcanopy column records which subset each image comes from, and the split column keeps the authors' 7:2:1 train/val/test split. The dataset contains 3,986 images…

Dataset · Object detection

sugar_beet_weed_detection

AgML

PaimioSB2023 is a field dataset for detecting sugar beet seedlings and weeds, collected in 2023 on a 2-hectare research sugar beet field in Hevonpää, Finland, where a commercial FarmDroid FD20 robot was evaluated for seeding and weeding. RGB images from a GoPro action camera mounted 2 m above the ground on the robot, a Parrot Anafi drone flying at 4 m, and a mobile phone were cut into 1280 x 1280 tiles, and each tile was manually annotated with bounding boxes for two classes: sugar beet and weed. The sourceimage column identifies the original photo each tile came from, so tiles from the same photo can be kept in the same split. The dataset contains 18,593 images with 338,171 bounding box…