- 3D files are available for download for both Color Depth Searches and PatchPerPix (PPP) searches
- LM data: aligned image stacks can be downloaded in the h5j format
- EM data: skeletons can be downloaded in the swc format
LM data
- Added 27,600 additional samples from the Annotator Gen1 MCFO data set, including 579 new lines. These samples were imaged with 20x and 63x objectives instead of 40x. See the FlyLight Gen1 MCFO website (https://gen1mcfo.janelia.org/cgi-bin/gen1mcfo.cgi) for all available images and citation guidance.
The rest is as in 2.2.0
LM data
- Added MCFO phase 2 lines
The rest is as in 2.2.0
See 2.2.0
EM data
- Hemibrain 1.2.1
LM data
- MCFO Phase 2
Notes
- See https://github.com/Kainmueller-Lab/PatchPerPix for information on the algorithm
EM data
- Hemibrain 1.2.1
LM data
-
MCFO GAL4 lines
-
Stable split GAL4 lines
-
Parameters for pre-computed color-depth matches between LM and EM data
- maskThreshold: 20
- targetThreshold: 20
- xyShift : 2
- mirrorMask: enabled
- pixColorFluctuation: 1.0
- pctPositivePixels: 1.0
Notes
- Added
neuronType
andneuronInstance
attributes to EM results JSON - Added
searchablePNG
attribute to results JSON
EM data
- Hemibrain 1.2.0
LM data
- MCFO GAL4 lines
Notes
- See https://github.com/Kainmueller-Lab/PatchPerPix for information on the algorithm
- Fixed the merge process. There was a bug which favored matches with higher pixel score while merging results from EM-vs-MCFO with the results from EM-vs-SGal4 instead of the matches with a higher normalized gradient score. This led to a wrong ranking of the final results.
- Hemibrain 1.1
- 32,777 EM bodies
-
MCFO GAL4 lines
- 175,533 segmented MCFO CDM
- 80,812 MCFO channels
- 27,144 MCFO samples
- 4,540 MCFO GAL4 lines
-
Stable split GAL4 lines
- 8,727 segmented SS CDM
- 3,045 SS channels
- 1,379 SS samples
- 653 SS lines
-
Parameters for pre-computed matches between LM and EM data
- maskThreshold: 20
- targetThreshold: 20
- xyShift : 2
- mirrorMask: enabled
- pixColorFluctuation: 1.0
- pctPositivePixels: 1.0
- Data format change - removed the attributes map
- Precomputed matches between Flylight Split GAL4 and MCFO vs Hemibrain 1.0.1
- Created the segmentation for Flylight Split GAL4 and MCFO libraries and the color depth search between hemibrain neurons and flylight was actually run using the segmented MIPs
- For Hemibrain neurons which cross the middle section also generated the flipped neuron by mirroring it and adding the original and also compared this against all flylight MIPs.
- For calculating the negative scores - gradient area gap and the area of regions with high expression - we selected the MIPs for the top 300 lines and we used precomnputed RGB masks generated with a radius of 20px