Visual representation of how Cell-TRACTR uses track queries to track cells and detect divisions.
<p>(A) Overview of Cell-TRACTR performing cell tracking, where a black arrow connecting two cells signifies a cell division event, as seen in frame t<sub>2</sub>. (B) In frame t<sub>0</sub>, the model performs object detection and locates all cells in the frame. In subs...
محفوظ في:
| المؤلف الرئيسي: | |
|---|---|
| مؤلفون آخرون: | |
| منشور في: |
2025
|
| الموضوعات: | |
| الوسوم: |
إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
|
| الملخص: | <p>(A) Overview of Cell-TRACTR performing cell tracking, where a black arrow connecting two cells signifies a cell division event, as seen in frame t<sub>2</sub>. (B) In frame t<sub>0</sub>, the model performs object detection and locates all cells in the frame. In subsequent frames, the output embeddings serve as track queries, which track the locations of the cells from frame to frame. An output embedding split with two colors indicates that the model predicted a cell division. A black output embedding indicates when a cell, represented by a track query, is lost, such as when it exits the field of view. (C) Each output embedding makes two predictions. Each prediction can be classified either as a cell or no object. The class labels determine which prediction is used. If only the first prediction for an output embedding is classified as a cell, then just the first prediction is used. We ignore predictions from output embeddings that predict no object for both class labels. When both predictions are classified as cells, the output embedding is labeled as a division.</p> |
|---|