AI-MED-AGH/Recruitment-Task-3

DeepWeeds - AI-MED AGH convenience mirror

This is a convenience mirror of the official DeepWeeds image archive and the upstream annotations pinned to a specific commit. original/images.zip is preserved unchanged; images are not extracted or duplicated here. models.zip from the source authors is deliberately not mirrored.

Dataset facts

• 17,509 in-situ images from Queensland, Australia.
• Nine classes: eight weed species plus Negative.
• The authors publish five folds, each with train, validation, and test CSV partitions under labels/.

1. Chinee apple
2. Lantana
3. Negative
4. Parkinsonia
5. Parthenium
6. Prickly acacia
7. Rubber vine
8. Siam weed
9. Snake weed

Candidate instruction

Perform your own literature search before selecting methods, baselines, and evaluation choices. Cite the sources that inform your work.

Provenance

• Original repository: https://github.com/AlexOlsen/DeepWeeds
• Paper: https://doi.org/10.1038/s41598-018-38343-3
• Official image archive: https://drive.google.com/file/d/1xnK3B6K6KekDI55vwJ0vnc2IGoDga9cj
• Annotation commit: https://github.com/AlexOlsen/DeepWeeds/tree/da084f62bcd1a2b0afeb8b81f1a27be3186391a1
• Retrieved: 2026-09-10

Images and annotations are licensed under CC BY 4.0. The upstream repository code is Apache-2.0 only; it does not change the data license. See SOURCE.md, LICENSE-DATA, and SHA256SUMS for details.