Currently working on...[keep updating]

Posted on

I am current working on four caffe models for detection, classification and segmentation.
This is served as an log file.

cifar

for different net: parameter+h -> mat
for diff image: layers

  • Models used by the VGG team in ILSVRC-2014
    VGG_ILSVRC_19_layers
    for classification, imagenet 1000 categories(OUTUPT:10X1000 prob)
    check the classification result for category

  • GoogLeNet GPU implementation from Princeton.
    ``
    for classification, imagenet 1000 categories

  • CNN Models for Salient Object Subitizing.
    AlexNet_SalObjSub

    - AlexNet_SalObjSub.caffemodel
./scripts/download_model_from_gist.sh 0585ed9428dc5222981f
http://www.cs.bu.edu/groups/ivc/data/SOS/AlexNet_SalObjSub.caffemodel

    any image

    • visuilization

  • Fully Convolutional Semantic Segmentation Models (FCN-Xs)
    fcn-32s-pascal
    for segmentation, pascal voc 20 categories

    **FCN-8s PASCAL**: three stream, 8 pixel prediction stride version
    ./scripts/download_model_from_gist.sh 91eece041c19ff8968ee
    http://dl.caffe.berkeleyvision.org/fcn-8s-pascal.caffemodel
**FCN-32s PASCAL**: three stream, 32 pixel prediction stride version
    ./scripts/download_model_from_gist.sh ac410cad48a088710872
    wget http://dl.caffe.berkeleyvision.org/fcn-32s-pascal.caffemodel
    • visuilization

  • Conditional Random Fields as Recurrent Neural Networks
    bvlc_reference_rcnn_ilsvrc13
    for segmentation, pascal voc 20 categories

    ./scripts/download_model_from_gist.sh a7e397abbda52c0b90323c23ab95bdeabee90a98
http://goo.gl/j7PrPZ

1
2
3
4
5
6
7
download model: 
	./scripts/download_model_from_gist.sh <gist_id> <>
	// download model metadata, archiecture, solver configuration
download .caffemodel file

# move things up a dirextory
find . -mindepth 1 -maxdepth 1 -exec mv -t.. -- {} +

Test on
PASCAL 20 categories:

Aeroplanes
Bicycles
Birds
Boats
Bottles
Buses
Cars
Cats
Chairs
Cows
Dining tables
Dogs
Horses
Motorbikes
People
Potted plants
Sheep
Sofas
Trains
TV/Monitors