画像ポリシーに3D形状知識を蒸留する形状教師アライメント
Aligning Policy Encoders with a Shape Teacher
展開時は画像のみを使うポリシーの訓練中に、事前学習済み3D形状表現(Point-MAE)を教師として画像パッチ特徴を整合させ、幾何構造を獲得させる手法を提案。LIBEROで収束を大幅に加速し成功率も向上。
詳しい要約
1. どんなもの?
2. 先行研究と比べてどこがすごい?
3. 技術・手法の肝は?
4. どうやって有効だと検証した?
5. 議論はある?
6. 次に読むべき論文は?
※ AIが要旨から生成した要約です。正確性は原文をご確認ください。
著者: Madelon Bernardy, Mohammad Mahdi Derakhshani, David M Knigge, Riccardo Valperga, Artem Moskalev
分類: cs.RO, cs.AI, cs.CV
原文アブストラクト
3D-aware representations can improve robotic manipulation by providing geometric information that is robust to changes in viewpoint, object position, and occlusion. However, using 3D inputs at deployment requires additional sensing and processing, making image-based policies more practical. We ask whether an image-based policy can acquire 3D shape information during training while retaining a purely image-based interface at deployment. We propose shape-teacher alignment, a training-time mechanism that uses a pretrained 3D shape representation as an alignment target for an image-based policy encoder. Specifically, we use Point-MAE features from individual objects as targets for corresponding image patches, transferring geometric structure to the policy without changing its deployment inputs or architecture. On LIBERO, shape-teacher alignment produces image representations that are more structured around object shape and corresponding parts, while substantially accelerating policy convergence and improving success rates. For pretrained policies, convergence is accelerated by 4.7x and 5.6x on the Object and Spatial suites, respectively, without requiring the shape teacher or depth input at deployment.