You need to download the following datasets first
-
ALOV dataset - The ALOV video dataset can be downloaded here: http://alov300pp.joomlafree.it/
-
Imagenet dataset - Download the DET dataset (47GB) and bounding box annotations (15MB) from Imagenet. Extract the training images and the annotations.
Once you extracted the dataset please make the following changes in train_tracker.sh:
IMAGENET_FOLDER: this folder should contain images and gt (annotations). The folder structure is as shown below.
| images | gt |
|---|---|
 |
 |
ALOV_FOLDER: this folder should contain images and gt (annotations). The folder structure is as shown below.
| images | gt |
|---|---|
 |
 |
INIT_CAFFEMODEL: Path to pretrained model
TRACKER_PROTO: Path to tracker.prototxt
SOLVER_PROTO: Path to solver.prototxt
Run the following command
bash train_tracker.sh