WasteSortNet Scanner

Detect every waste item, then measure the mix two ways: by item count and by box area Loading ...

1 · Choose a photo

Drop a photo here or tap to upload or take a photo

Your photo is scanned inside this browser and is not uploaded anywhere.

Test photos

Held-out TACO test photos: the model never saw them during training.

2 · Detections (PO1)

No photo yetUpload a photo, use the camera, or pick a test photo.
Photo with the detected waste items
Scanning ...

3 · Composition of this photo (PO2)

By item countBy box area

Count and area can disagree

By count, the photo is mostly
By area, the photo is mostly

How far the two measures disagree (0 = same answer, 2 = completely different)

Checked against the labels

Count error (L1)
Area error (L1)

Detected items

    WasteSortNet, a Final Year Project by Taufik Anuar, B.Sc. Data Science, Universiti Malaysia Sabah. Detector: YOLOv8s (Ultralytics), fine-tuned on TACO, test mAP@0.5 0.219. Test photos: TACO dataset (Proença & Simões, 2020), CC BY 4.0.