Heatmap illustrating which multi-scale feature pixels were classified as “cell” by query selection. This data was collected by processing the full test set for the DeepCell dataset. Note that for these movies, Cell-TRACTR preferentially uses pixels from the highest resolution multi-scale features.

مواد مشابهة

• Heatmap illustrating which multi-scale feature pixels were classified as “cell” by query selection. This data was collected by processing the full test set containing 29 movies from the bacterial mother machine dataset with Cell-TRACTR. Note that for these movies, the model preferentially uses pixels from the lowest resolution multi-scale features. The black pixel at the top corresponds to the top of the mother machine image where there are not usually cells.
  حسب: Owen M. O’Connor (11965460)
  منشور في: (2025)
• Time-lapse microscopy movie from the mammalian DeepCell test set processed by Cell-TRACTR.
  حسب: Owen M. O’Connor (11965460)
  منشور في: (2025)
• (A) Example illustrating the correlation between the final prediction of an object query and where the object query is derived from within the multi-scale features. Colored pixels represent multi-scale features that were classified as “cell” by the decoder. Black pixels represent multi-scale feature pixels that were classified as “no object” by the decoder. The colored pixels are overlaid onto the phase contrast image for visualization purposes. (B) A more complex example illustrating the correlation between the multi-scale features and object queries.
  حسب: Owen M. O’Connor (11965460)
  منشور في: (2025)
• Evaluation of model performance on the mammalian DeepCell dataset.
  حسب: Owen M. O’Connor (11965460)
  منشور في: (2025)
• Depiction of how predictions are made from the output embeddings. The largest encoded multi-scale feature map is resized and combined with the original multi-scale features to form the pixel embedding map. The dot product of the output embeddings and the pixel embedding map produces the segmentation masks. A linear layer is used to predict the class label and a multi-layer perceptron (MLP) is used to generate the bounding box. For output embeddings classified as “no object”, the bounding boxes and segmentation masks are ignored.
  حسب: Owen M. O’Connor (11965460)
  منشور في: (2025)