Closed drinkingkazu closed 6 years ago
Yes, I've reviewed APICaffe in LArCV and your caffe branch which loads via the thread factory. I don't know your exact configuration for SimpleFiller (maybe you can point me to it) but I assume I can let pycaffe create a ROOTLayer then access the instantiated SimpleFiller and call fill_entry_data bypassing the need for a ROOT file on disk.
Given ROOT data layer caffe config looking like this:
layer { name: "data" type: "ROOTData" top: "data" top: "label"
root_data_param {
batch_size: 1
filler_config: "ana_filler.cfg"
filler_name: "Analysis"
}
}
... to process 1 event at a time, here's how filler config looks like
Analysis: { Verbosity: 2 EnableFilter: false RandomAccess: false
InputFiles: ["/stage/drinkingkazu/production/v03/hires_crop/data_extbnb/part01/hiresdiv_extbnb_0000_0009.root","/stage/drinkingkazu/production/v03/hires_crop/data_extbnb/part01/hiresdiv_extbnb_0010_0019.root","/stage/drinkingkazu/production/v03/hires_crop/data_extbnb/part01/hiresdiv_extbnb_0020_0039.root","/stage/drinkingkazu/production/v03/hires_crop/data_extbnb/part01/hiresdiv_extbnb_0040_0059.root","/stage/drinkingkazu/production/v03/hires_crop/data_extbnb/part01/hiresdiv_extbnb_0060_0079.root","/stage/drinkingkazu/production/v03/hires_crop/data_extbnb/part01/hiresdiv_extbnb_0080_0099.root","/stage/drinkingkazu/production/v03/hires_crop/data_extbnb/part01/hiresdiv_extbnb_0100_0119.root","/stage/drinkingkazu/production/v03/hires_crop/data_extbnb/part01/hiresdiv_extbnb_0140_0159.root","/stage/drinkingkazu/production/v03/hires_crop/data_extbnb/part01/hiresdiv_extbnb_0160_0180.root"]
ProcessType: ["SimpleFiller"] ProcessName: ["SimpleFiller"]
ProcessList: { SimpleFiller: {
MeanADC: [0,0,0]
MeanImageFile: "/stage/drinkingkazu/production/v03/hires_crop/data_extbnb/mean_part01.root"
MeanImageProducer: "mean"
InputProducer: "tpc_hires_crop"
ROIProducer: "tpc_hires_crop"
# SimpleFiller configuration
Channels: []
MaxADC: [200,200,200]
MinADC: [17,13,20]
GausSmearingMean: 1.0
GausSmearingSigma: 0.1
PixelWiseSmearing: true
}
} }
I think this is a legacy feature so I would close. But please re-open if needed!
We want to run the viewer with exactly same way image is passed onto the network training and/or forward computation. New branch: https://github.com/LArbys/caffe/tree/process_driver/src/caffe ... has the implementation we planed already in place.
need this to be implemented in view_rgb.py