Update README.md

README.md

@@ -40,12 +40,16 @@ conda install matplotlib
 
 ## Training/Evaluation on Visual Genome
 If you want to **train/evaluate** RelTR on Visual Genome, you need a little more preparation:
+a) Scipy (we used 1.5.2) is required. 
+```
+conda install scipy
+```
 
-a) Download the annotations of [Visual Genome (in COCO-format)](https://drive.google.com/file/d/1aGwEu392DiECGdvwaYr-LgqGLmWhn8yD/view?usp=sharing) and unzip it in the ```data/``` forder.
+b) Download the annotations of [Visual Genome (in COCO-format)](https://drive.google.com/file/d/1aGwEu392DiECGdvwaYr-LgqGLmWhn8yD/view?usp=sharing) and unzip it in the ```data/``` forder.
 
-b) Download the the images of VG [Part1](https://cs.stanford.edu/people/rak248/VG_100K_2/images.zip) and [Part2](https://cs.stanford.edu/people/rak248/VG_100K_2/images2.zip). Unzip and place all images in a folder ```data/vg/images/```
+c) Download the the images of VG [Part1](https://cs.stanford.edu/people/rak248/VG_100K_2/images.zip) and [Part2](https://cs.stanford.edu/people/rak248/VG_100K_2/images2.zip). Unzip and place all images in a folder ```data/vg/images/```
 
-c) Some widely-used evaluation code (**IoU**) need to be compiled... We will replace it with Pytorch code.
+d) Some widely-used evaluation code (**IoU**) need to be compiled... We will replace it with Pytorch code.
 ```
 # compile the code computing box intersection
 cd lib/fpn