Pre-trained PlantCLEF 2019 models and supplementary material

In order to support research in fine-grained plant classification and to allow full reproducibility of our results, we share the pre-trained Inception-v4 and Inception-ResNet-v2 CNN models trained for the PlantCLEF 2019 plant identification task, along with changes to the training data. In the post-challenge evaluation, the method achieves the best results on the PlantCLEF 2019 test set.

Usage

The models are shared in the form of TensorFlow checkpoints, and can be used directly within the TensorFlow slim framework. Information about the model parameters are described in our LifeCLEF 2019 paper (awaiting publication).

Models

The models listed below were fine-tuned from our previous year's checkpoints shared with other participants of the 2019 task.

[DIR] inception_resnet_v2/
[DIR] inception_resnet_v2_second/
[DIR] inception_v4/
[DIR] inception_v4_second/
[DIR] inception_v4_x2/



Data

- List of detected noisy images, which we removed from the official training set.
- Archive of additional images used for training, which were downloaded from GBIF. The images were tested to have zero overlap with the PlantCLEF 2019 test set.

License

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
I.e.: The models are shared only for non-commercial purposes. If you publish experiments/results based on the models, please attribute us by citing the related paper, Recognition of the Amazonian flora by Inception Networks with Test-time Class Prior Estimation. BibTex:
@InProceedings{cmp2019lifeclef,
  title={Recognition of the Amazonian flora by Inception Networks with Test-time Class Prior Estimation},
  author={Picek, Lukas and Sulc, Milan and Matas, Jiri},
  booktitle= {Working Notes of CLEF 2019 - Conference and Labs of the
		  Evaluation Forum},
  year={2019}
}