Skip to content
Keras implementation of Deeplab v3+ with pretrained weights
Branch: master
Clone or download
Latest commit c683d71 Mar 19, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
LICENSE Create LICENSE Apr 4, 2018 add MobileNetv2 backbone May 10, 2018

Keras implementation of Deeplabv3+

DeepLab is a state-of-art deep learning model for semantic image segmentation.

Model is based on the original TF frozen graph. It is possible to load pretrained weights into this model. Weights are directly imported from original TF checkpoint.

Segmentation results of original TF model. Output Stride = 8

Segmentation results of this repo model with loaded weights and OS = 8
Results are identical to the TF model

Segmentation results of this repo model with loaded weights and OS = 16
Results are still good

How to get labels

Model will return tensor of shape (batch_size, height, width, num_classes). To obtain labels, you need to apply argmax to logits at exit layer. Example of predicting on image1.jpg:

from matplotlib import pyplot as plt
import cv2 # used for resize. if you dont have it, use anything else
import numpy as np
from model import Deeplabv3
deeplab_model = Deeplabv3()
img = plt.imread("imgs/image1.jpg")
w, h, _ = img.shape
ratio = 512. / np.max([w,h])
resized = cv2.resize(img,(int(ratio*h),int(ratio*w)))
resized = resized / 127.5 - 1.
pad_x = int(512 - resized.shape[0])
resized2 = np.pad(resized,((0,pad_x),(0,0),(0,0)),mode='constant')
res = deeplab_model.predict(np.expand_dims(resized2,0))
labels = np.argmax(res.squeeze(),-1)

How to use this model with custom input shape and custom number of classes

from model import Deeplabv3
deeplab_model = Deeplabv3(input_shape=(384,384,3), classes=4)  

After that you will get a usual Keras model which you can train using .fit and .fit_generator methods.

How to train this model

You can find a lot of useful parameters in the original repository.

Important notes:

  1. This model don't have default weight decay, you need to add it yourself;
  2. Xception backbone should be trained with OS=16, and only inferenced with OS=8;
  3. You can freeze feature extractor for Xception backbone (first 356 layers) and only fine-tune decoder;
  4. If you want to train BN layers too, use batch size of at least 12 (16+ is even better).

Known issues

As far as we know, this model can't be fine tuned as is and is only usable for inference. See this issue for a discussion around this and possible alternatives.

Don't hesitate to discuss or submit a pull request if you've got ideas on how to fix this model.

How to load model

In order to load model after using use this code:

from model import relu6, BilinearUpsampling
deeplab_model = load_model('example.h5',custom_objects={'relu6':relu6,'BilinearUpsampling':BilinearUpsampling })

Xception vs MobileNetv2

There are 2 available backbones. Xception backbone is more accurate, but has 25 times more parameters than MobileNetv2.

For MobileNetv2 there are pretrained weights only for alpha=1. However, you can initiate model with different values of alpha.

Requirement (it may work with lower versions too, but not guaranteed)


You can’t perform that action at this time.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session.