Person re-identification (Re-ID) aims to identify the same Person from a variety of non-overlapping viewpoints from multiple cameras. In this project, we present a distance metric-based deep learning framework capable of Person Re-ID. A deep cosine metric learning model is kept as the baseline of our framework which uses a CNN architecture with residual blocks (BigNet) to generate feature map and uses the cosine softmax classifier during training. The baseline without augmented dataset gives a maximum mAP score of 0.861 on the provided validation set. Training the model on augmented dataset improved the maximum mAP score to 0.878. The baseline model was improved by considering a horizontal slicing of images (SliceNet1) before passing it to BigNet and the obtained features were concatenated before passing it to the cosine softmax classifier. A variation of the above model (SliceNet2) considered the slicing of feature map generated by a modified BigNet and concatenating them before passing it to cosine softmax classifier. A maximum mAP score of 0.951 and 0.953 was observed for SliceNet1 and SliceNet2 respectively.