diff --git a/readme.md b/readme.md
index b3ed8c376..2eb5d9144 100644
--- a/readme.md
+++ b/readme.md
@@ -82,7 +82,7 @@ The speed/FPS test includes the time of post-processing with no jit/data precisi
 ## Demo
 
     # install requirements
-    pip install pycocotools numpy opencv-python tqdm tensorboard tensorboardX pyyaml webcolors
+    pip install pycocotools numpy opencv-python tqdm tensorboard tensorboardX pyyaml webcolors wandb
     pip install torch==1.4.0
     pip install torchvision==0.5.0
      
diff --git a/train.py b/train.py
index 04133cb0c..0e9780f97 100644
--- a/train.py
+++ b/train.py
@@ -10,6 +10,7 @@
 import numpy as np
 import torch
 import yaml
+import wandb
 from tensorboardX import SummaryWriter
 from torch import nn
 from torch.utils.data import DataLoader
@@ -173,6 +174,7 @@ def freeze_backbone(m):
     else:
         use_sync_bn = False
 
+    wandb.init() # initialize a wandb run
     writer = SummaryWriter(opt.log_path + f'/{datetime.datetime.now().strftime("%Y%m%d-%H%M%S")}/')
 
     # warp the model with loss function, to reduce the memory usage on gpu0 and speedup
@@ -248,7 +250,13 @@ def freeze_backbone(m):
                     # log learning_rate
                     current_lr = optimizer.param_groups[0]['lr']
                     writer.add_scalar('learning_rate', current_lr, step)
-
+                    wandb.log({"step" : step,
+                                "epoch": epoch,
+                                "train-loss": loss,
+                                "train-reg-loss": reg_loss,
+                                "train-cls-loss" : cls_loss,
+                                "lr" : current_lr
+                                })
                     step += 1
 
                     if step % opt.save_interval == 0 and step > 0:
@@ -296,6 +304,12 @@ def freeze_backbone(m):
                 writer.add_scalars('Regression_loss', {'val': reg_loss}, step)
                 writer.add_scalars('Classfication_loss', {'val': cls_loss}, step)
 
+                wandb.log({"step" : step,
+                            "val-loss" : loss,
+                            "val-reg-loss" : reg_loss,
+                            "val-cls-loss" : cls_loss,
+                            "epoch" : epoch})
+
                 if loss + opt.es_min_delta < best_loss:
                     best_loss = loss
                     best_epoch = epoch