Ar Pusht Image: Arabic-Translated Vision-Language-Action Robotics Dataset

An Arabic-translated version of the PushT image dataset for robotics. The dataset provides high-quality Arabic instructions for complex manipulation tasks, enabling the training of localized robotics policies. It was created by hamzabouajila and last updated on 2026-04-21.

Use Cases

• Train localized robotics policies based on native Arabic instructions.
• Benchmark Vision-Language-Action models on Arabic manipulation tasks.
• Fine-tune models like FastVLA and OpenVLA for Arabic-language robotics.
• Study the impact of language localization on robotic instruction following.

Strengths

• First Arabic-translated dataset for Vision-Language-Action robotics.
• High-quality translations for complex manipulation tasks.
• Compatible with established VLA frameworks like FastVLA and OpenVLA.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Localized version of the standard PushT image dataset.

Collection Method

Translation of every instruction into Arabic.

Time Range

null

Freshness

Last updated 2026-04-21 16:54:37; freshness should be verified.

Geography

null