LeRobot ACT · Tridex 3Cam ResNet 100K 18_11_2

Community implementation
Papers and technical reports

Tridex_DIA_Niort trained this LeRobot ACT policy to pick up a highlighter and place it in a pink box. The public ResNet-18 checkpoint uses front, side and top RGB images plus six robot-state values and predicts six joint and gripper commands per step.

View full model description

Papers and technical reports

Paper / report unconfirmed

Official introductory materials are recorded; no version-specific paper or technical report has been verified.

Last checked: 2026-11-21

How to access and use

Model files
Public access
Training data
Public access

Training code for this version is unconfirmed. Public code resources do not establish access to the exact training implementation. View verification notes

Training / fine-tuning code

Model files

Model files
LeRobot ACT · Tridex 3Cam ResNet 100K 18_11_2 · 6ba88ecAccessWeights license
LeRobot ACT · Tridex 3Cam ResNet 100K 18_11_2AccessWeights license
Release information

SourceVerified on · 2026-11-21

Checks before deployment

The dataset uses the shared SO follower identifier (so_follower). The exact arm variant, policy control rate and evaluation results for these weights are unconfirmed. Available weights or code do not establish plug-and-play robot compatibility. Check the required configuration, data and interfaces in the source documentation.

Deployment and usage resources

Dataset format
LeRobot v3.0
Runtime and frameworks
LeRobot

Architecture and inputs / outputs

Model type
Action policy
Architecture
Transformer
Backbone
ResNet-18
Weight parameters
51668614 parameters
Input
RGB images / Robot state
Released
November 21, 2026

Scope and training

joint_position · continuous

6 DoF · horizon 100 · mean_std

Functional purpose

action_control

Record identity

Community implementation

Output

Actions

Pipeline task

robotics

Robot form

Robotic arm

Task scope

Single task

Training data scale

33 episodes

Training

Imitation learning

Lineage

Derived from: ACT · LeRobot ACT ↗

Related robots

Full parameters and verification notes (model card)

The policy takes three 640 × 480 RGB views and a six-dimensional state, and predicts 100-step chunks of six-dimensional actions. The publisher lists the weights under Apache-2.0.

The LeRobot 0.6.2 configuration specifies 100,000 training steps, AdamW, a batch size of 8, a learning rate of 0.00001 and seed 1000. It selects episodes 0–32 from the OVERLAY dataset, which contains 38 demonstrations and 16,424 frames recorded at 30 FPS.

Training dataset

Submit a source or correction