thtrieu / darkflow

Translate darknet to tensorflow. Load trained weights, retrain/fine-tune using tensorflow, export constant graph def to mobile devices
GNU General Public License v3.0
6.13k stars 2.08k forks source link

create annotations #277

Open hrshovon opened 7 years ago

hrshovon commented 7 years ago

I am new to yolo. I am trying to train a model to detect numbers in a image. I am using MNIST dataset. So far I have a folder called training data where the 28x28 images are in their respective folders(like all images of 5 are in folder named 5 etc.) I have created a labels.txt file which contains list of classes(0 to 9) I have copied tiny-yolo.cfg in my folder and modified classes and filters variables. Now what to do about annotations.How do I create that file and how do I initiate the weight file?

bauerzhou commented 7 years ago

I used labelImg to label my own dataset.

wendq86 commented 7 years ago

I organised my images and annotations the same way as the Pascal VOC dataset and used labelImg for generating the annotation file for each image. You can load the weight file via the option --load bin/yolo-tiny.weights

jubjamie commented 7 years ago

I've been converting my own dataset format to VOC today and it's not too bad. But the above tool is great for starting from scratch with images! Nice find @bauerzhou

hardikchauhan8 commented 6 years ago

How can I create Annotation xmls, can anyone help me with this?

finvernizzi commented 6 years ago

labelimg directly generates the XMLs files you need