By Lachlan Tychsen-Smith (research engineer), Lars Petersson (supervisor) https://arxiv.org/abs/1703.10295
- DeNet-34 skip: Google Drive
- DeNet-101 skip: Google Drive
Use the following command to evaluation models on gpu0:
GPU=0 model-predict --batch-size 8 --thread-num 4 --predict-mode detect,voc --input VOC_DIR --model MODEL_FNAME --extension voc,2007-test,images_per_subset=128,crop=512,scale_mode=large
where:
- VOC_DIR is the path to the VOCdevkit directory
- MODEL_FNAME is the path to the downloaded model file (see above)
To output images with annotated bounding boxes set "--predict-mode detect,voc,image"