- Mindspore=2.0.0
- Convolutional Neural Networks
- Long Short Term Memory Cells
Clone the repo:
git clone https://openi.pcl.ac.cn/Kayxxx/ShowAndTell.git
- Prepare Dataset (Flickr8k).
- Extract and move images to a folder named Images and text to captions.txt.
- Put the folder containing Images and captions.txt in a folder named flickr8k
-- flickr8k
|-- Images
|-- 1000268201_693b08cb0e.jpg
|-- ......
|-- captions.txt
- Run the following command :
python train.py
- Run the following command :
python inference.py --image_path <path_to_image> --model_path <path_to_model>
assets
Some of the results obtained are shown below :
Caption : a man is standing on top of a mountain gazing at the sunset .
Bad Case:
Caption : a young boy is holding a yellow ball.