

Draw a schematic diagram of "ECN" (Expected Contact Network) for dexterous hand grasping, in academic conference style (ICRA/RSS). - Input: "Hand + Object Point Cloud per Frame" (hand blue, object orange), annotated "N_h x 3" / "N_o x 3".- Arrow → "Point Cloud Encoder" (PointNet / DGCNN), outputs per-frame feature h_t.- Arrow → "Temporal Model" (Transformer / TCN / LSTM), sequence of h_t → z_t, capturing time dependencies, show time axis.- Arrow → "Contact Decoder" (multi-head MLP), outputs per-frame: s_t^i (contact probability), df_t^i (contact force change rate).- Final arrow → Output sequence [s_t^i, df_t^i] for t=1..H.- Style: clear, minimalistic, professional, boxes and arrows, labels readable, color-coded for clarity.
描述你想生成的图片信息,如:画面里有什么风格/元素/色彩
Shift+回车 = 换行
AI可能会犯错误或不完全遵循,请谨慎使用。
同款 0
评论 0