| ... | ... | @@ -14,11 +14,12 @@ Discussion pictures: |
|
|
|
|
|
|
|
Notable outcomes:
|
|
|
|
|
|
|
|
* Kate shared that Segmentaiton is much better 16-bit vs 8-bit on deepcell
|
|
|
|
* Implement k-NN metric function for monitoring
|
|
|
|
* Adding special mask channels with bias information
|
|
|
|
* Add channels of cell boundaries from DeepCell
|
|
|
|
* Add cell mask channel
|
|
|
|
* Foveated Attention models? <https://www.semanticscholar.org/paper/Peripheral-Vision-Transformer-Min-Zhao/6fb05e3483d05f79e59e5ce957708d8ab8932fdc>
|
|
|
|
* Foveated Attention models? <https://www.semanticscholar.org/paper/Peripheral-Vision-Transformer-Min-Zhao/6fb05e3483d05f79e59e5ce957708d8ab8932fdc>
|
|
|
|
* Implement H&E connection to model
|
|
|
|
* remove bad channels
|
|
|
|
* try H&E + "action" markers fine-tuned.
|
| ... | ... | |