Kaggle Competition Project as well as ANLY 590 Final Project.
This repo is the solution for Kaggle Competition Plant Seedlings Classification as well as the final project of ANLY 590.
Task: Determine the species of a seedling from an image.
More information related to project could be found at Project Proposal.
Poster could be found here.
Final Paper could be found here.
You could download data from website or use Kaggle API as following:
cd input
kaggle competitions download -c plant-seedlings-classificationPlease download data into input folder.
pip install -r requirements.txt# In Plant_Seedlings_Classification Folder
jupter notebookYou could check notebook via Jupter.
| Method Used | Parameters | F1 |
|---|---|---|
| VGG19 | 20,270,264 | 0.80 |
| 6 Layers | 3,320,396 | 0.93306 |
| 6 Layers + Background Remove | 3,320,396 | 0.95843 |
| 9 Layers + Fine Tuning | 1,775,100 | 0.97103 |
| 9 Layers + Background Remove | 1,775,100 | 0.97507 |
| 9 Layers + ADASYN | 1,775,100 | 0.97799 |
| 9 Layers + Random Up Sample | 1,775,100 | 0.97744 |
| 9 Layers + SMOTE | 1,775,100 | 0.97832 |
| 9 Layers + ADASYN + Data Augmentation | 1,775,100 | 0.97902 |
| 9 Layers + ADASYN + Data Augmentation + Snapshot Ensemble | 1,775,100 | 0.98740 |
| CNN DenseNet 121 + Data Augmentation + Background Remove | 8,062,504 | 0.94710 |
| CNN DenseNet 121 + Data Augmentation + GAN + Background Remove | 8,062,504 | 0.97984 |
| State of Art on Kaggle[8] | 54,521,176 | 0.99496 |
[1] Thomas Mosgaard Giselsson and (2017). A Public Image Database for Benchmark of Plant Seedling Classification. CoRR, abs/1711.05458.
[2] Giselsson, Thomas Mosgaard, Dyrmann, Mads, Jorgensen, Rasmus Nyholm, Jensen, Peter Kryger & Midtiby, Henrik Skov (2017). A Public Image Database for Benchmark of Plant Seedling Classification Algorithms.
[3] J. Schwiegerling, Field Guide to Visual and Ophthalmic Optics, SPIE Press, Bellingham, WA (2004).
[4] Haibo He, Yang Bai, E. A. Garcia and Shutao Li, "ADASYN: Adaptive synthetic sampling approach for imbalanced learning," 2008 IEEE International Joint Conference on Neural Networks (IEEE World Congress on Computational Intelligence), Hong Kong, 2008, pp. 1322-1328. doi: 10.1109/IJCNN.2008.4633969
[5] Chawla, N. V. et al. “SMOTE: Synthetic Minority Over-Sampling Technique.” Journal of Artificial Intelligence Research 16 (2002): 321–357. Crossref. Web.
[6] Huang, G., Li, Y., Pleiss, G., Liu, Z., Hopcroft, J. E., & Weinberger, K. Q. (2017). Snapshot ensembles: Train 1, get M for free. arXiv preprint arXiv:1704.00109 .
[7] Guillaume Lemaitre, Fernando Nogueira & Christos K. Aridas (2017). Imbalanced-learn: A Python Toolbox to Tackle the Curse of Imbalanced Datasets in Machine Learning. Journal of Machine Learning Research, 18, 1-5.
[8] Kumar Shridhar., Kaggle #1 Winning Approach for Image Classification Challenge. Web Blog Post, Neural Space, June 20, 2018.
[9] Karl Pearson, 1901 K. Pearson, "On lines and planes of closest fit to systems of points in space", The London, Edinburgh and Dublin Philosophical Magazine and Journal of Science, Sixth Series, 2, pp.559-572 (1901) (1857-1936)
[10] L.J.P. van der Maaten and G.E. Hinton. Visualizing Data Using t-SNE. Journal of Machine Learning Research 9(Nov):2579-2605,2008.
[11] Weiss, Karl, et al. “A Survey of Transfer Learning.” Journal of Big Data , vol. 3, no. 1, 2016.
[12] Hinterstoisser, Stefan, et al. “On Pre-Trained Image Features and Synthetic Images for Deep Learning.” 2017.
[13] Wu, Cinna, Mark Tygert, and Yann LeCun. "Hierarchical loss for classification." arXiv preprint arXiv:1709.01062 (2017). 5 Radford A, Metz L, Chintala S. Unsupervised representation learning with deep convolutional generative adversarial networks[J]. arXiv preprint arXiv:1511.06434, 2015.
[14] Gao F, Yang Y, Wang J, et al. A deep convolutional generative adversarial networks (DCGANs)-based semi-supervised method for object recognition in synthetic aperture radar (SAR) images[J]. Remote Sensing, 2018, 10(6).