2. Export Artifacts

When you export your project for model learning and testing, the following artifacts are generated. Note that <dir> is a placeholder and refers to the directory name to which you have exported.

Output	Type	Description	darkflow	darknet
images	directory	stores images	yes	yes
<dir>_labels.txt	file	stores classes	yes	yes
annots	directory	stores XML annotations	yes	no
darkflow-train.sh	file	training script	yes	no
darkflow-test.sh	file	testing script	yes	no
tiny-yolo-<dir>.cfg	file	network architecture	yes	no
tiny-yolo-4c-<dir>.cfg	file	network architecture	yes	no
tiny-yolo-voc-<dir>.cfg	file	network architecture	yes	no
yolo-<dir>.cfg	file	network architecture	yes	no
yolo-voc-<dir>.cfg	file	network architecture	yes	no
labels	directory	stores TXT annotations	no	yes
darknet-train.sh	file	training script	no	yes
darknet-test.sh	file	testing script	no	yes
<dir>.data	file	training metadata	no	yes
<dir>_train.txt	file	list of training images	no	yes
<dir>_valid.txt	file	list of validation images	no	yes
yolov3-tiny-<dir>.cfg	file	network architecture	no	yes

As shown by the table above, the output artifacts are directories and files, and, some of these directories and files are used by darkflow, darknet or both. IAIA is attempting to make your deep learning object detection training and testing experience as easy as possible by generating all these outputs and organizing them in this way. In the end, all you have to do is to run the training and testing scripts to learn and validate your model using darkflow or darknet.