SAM3 Auto-Labeling

Generate segmentation masks and bounding boxes from plain-text prompts using Meta's Segment Anything 3. Build high-quality training sets with minimal redundancy — especially useful for video and large-scale datasets.

🎯

Auto-label

SAM3 predicts masks/boxes for every image using your text prompts

⚖

Zero-weight start

All samples start with weight = 0 — nothing is included until you accept it

📊

Curate

Decimate redundant frames, review clusters, catch wrong predictions in the Dashboard

🚀

Train

Accepted samples get weight = 1 — best model with the least data

1 Setup labels

2 Preview results

3 Create table & predict

Config

1 Setup & Preview

Config Name

Image Source

Folder Path

Table URL

Point to an image folder or existing 3LC table. SAM3 will process every image in the source.

Labels (text prompts)

Define the object classes you want to detect. Use descriptive text prompts that SAM3 can understand (e.g. "fish", "person walking").

Modality

Confidence: 0.20

2 Create Table

Create a 3LC table from the image folder. All samples initialized with weight = 0.

Project

Dataset Name

Table Name

This creates a new 3LC table with bounding boxes and segmentation masks for each detected object.

3 Run Prediction

Run SAM3 on every image. Labels read from table schema. Creates a 3LC Run with UMAP embeddings.

Table URL

Embeddings

Device

Progress

No active job.

Log

Preview

🔍 Add labels and click
Preview Random Image