minor update to look good visually

app.py

@@ -100,7 +100,11 @@ def sepia(input_img):
 title = "SegFormer(ADE20k) in TensorFlow"
 description = """
 
-This is demo TensorFlow SegFormer from 🤗 `transformers` official package. The pre-trained model was trained to segment scene specific images. We are **currently using ONNX model converted from the TensorFlow based SegFormer to improve the latency**. The average latency of an inference is **21** and **8** seconds for TensorFlow and ONNX converted models respectively (in [Colab](https://github.com/deep-diver/segformer-tf-transformers/blob/main/notebooks/TFSegFormer_ONNX.ipynb)). Check out the [repository](https://github.com/deep-diver/segformer-tf-transformers) to find out how to make inference, finetune the model with custom dataset, and further information.
+This is demo TensorFlow SegFormer from 🤗 `transformers` official package. 
+The pre-trained model was trained to segment scene specific images. We are **currently using ONNX model converted from the TensorFlow based SegFormer to improve the latency**. 
+The average latency of an inference is **21** and **8** seconds for TensorFlow and ONNX converted models respectively 
+(in [Colab](https://github.com/deep-diver/segformer-tf-transformers/blob/main/notebooks/TFSegFormer_ONNX.ipynb)).
+Check out the [repository](https://github.com/deep-diver/segformer-tf-transformers) to find out how to make inference, finetune the model with custom dataset, and further information.
 
 """