Update README.md

Author: amiralush

README.md

@@ -8,29 +8,25 @@ From the authors of [GraphLab](https://github.com/jegonzal/PowerGraph) and [Turi
 ![alt text](https://github.com/visualdatabase/fastdup/blob/main/gallery/mscoco_duplicates-min.png)
 *Duplicates and near duplicates identified in [MS-COCO](https://cocodataset.org/#home) and [Imagenet-21K](https://www.image-net.org) dataset*
 
-![alt text](https://github.com/visualdatabase/fastdup/blob/main/gallery/landmark_outliers-min.png)
-*Outliers in the [Google Landmark Recognition 2021 dataset](https://www.kaggle.com/competitions/landmark-recognition-2021) (dataset intention is to capture recognizable landmarks, like the empire state building etc.)*
-
 ![alt text](https://github.com/visualdatabase/fastdup/blob/main/gallery/imdb_outliers-min.png)
 *[IMDB-WIKI](https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/ ) outliers (data goal is for face detection, gender and age detection)*
 
+![alt text](https://github.com/visualdatabase/fastdup/blob/main/gallery/landmark_outliers-min.png)
+*Outliers in the [Google Landmark Recognition 2021 dataset](https://www.kaggle.com/competitions/landmark-recognition-2021) (dataset intention is to capture recognizable landmarks, like the empire state building etc.)*
+
 ![alt text](https://github.com/visualdatabase/fastdup/blob/main/gallery/red_wine.png)
 *Cluster of wrong labels in the [Imagenet-21K](https://www.image-net.org) dataset. No human can tell those red wine flavors from their image.*
 
-![alt text](https://github.com/visualdatabase/fastdup/blob/main/gallery/daisy.png)
-*Cluster of wrong labels in the [Imagenet-21K](https://www.image-net.org) dataset.*
-
-
-
-
 ![alt text](https://github.com/visualdatabase/fastdup/blob/main/gallery/imagenet21k_wrong_labels-min.png)
-*Wrong labels in the [Imagenet-21K](https://www.image-net.org) dataset*
+*Wrong labels in the [Imagenet-21K](https://www.image-net.org) dataset* Different labels to visaully similar daisy flower images.
+
+![alt text](https://github.com/visualdatabase/fastdup/blob/main/gallery/daisy.png)
+*Cluster of wrong labels in the [Imagenet-21K](https://www.image-net.org) dataset.* Different labels to visually similar red-wine images.
 
 ![alt text](https://github.com/visualdatabase/fastdup/blob/main/gallery/imagenet21k_funny-min.png)
 *Fun labels in the [Imagenet-21K](https://www.image-net.org) dataset*
 
 
-
 ## Results on Key Datasets
 We have thoroughly tested fastdup across various famous visual datasets. Ranging from pilar Academic datasets to Kaggle competitions. A key finding we have made using FastDup is that there are ~1.2M (!) duplicate images on the ImageNet-21K dataset, out of which 104K pairs belong both to the train and to the val splits (this amounts to 20% of the validation set). This is a new unknown result! Full results are below. * train/val splits are taken from https://github.com/Alibaba-MIIL/ImageNet21 .