Added the Cuda support for pytorch implementation

DenseDepth.ipynb

@@ -3,8 +3,6 @@
   "nbformat_minor": 0,
   "metadata": {
     "colab": {
-      "name": "DenseDepth",
-      "version": "0.3.2",
       "provenance": [],
       "collapsed_sections": [],
       "include_colab_link": true
@@ -23,24 +21,23 @@
         "colab_type": "text"
       },
       "source": [
-        "<a href=\"https://colab.research.google.com/github/ialhashim/DenseDepth/blob/master/DenseDepth.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
+        "<a href=\"https://colab.research.google.com/github/Avi241/DenseDepth/blob/master/DenseDepth.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
       ]
     },
     {
       "cell_type": "code",
       "metadata": {
         "id": "ahkR4C5dEnR0",
-        "colab_type": "code",
         "colab": {
           "base_uri": "https://localhost:8080/",
           "height": 139
         },
         "outputId": "491e7243-6dee-4e32-95c9-a817004775cd"
       },
       "source": [
-        "!git clone https://github.com/ialhashim/DenseDepth.git"
+        "!git clone https://github.com/Avi241/DenseDepth.git"
       ],
-      "execution_count": 1,
+      "execution_count": null,
       "outputs": [
         {
           "output_type": "stream",
@@ -61,7 +58,6 @@
       "cell_type": "code",
       "metadata": {
         "id": "fFQgwMlNExak",
-        "colab_type": "code",
         "colab": {
           "base_uri": "https://localhost:8080/",
           "height": 208
@@ -71,7 +67,7 @@
       "source": [
         "!wget https://s3-eu-west-1.amazonaws.com/densedepth/nyu.h5 -O ./DenseDepth/nyu.h5"
       ],
-      "execution_count": 2,
+      "execution_count": null,
       "outputs": [
         {
           "output_type": "stream",
@@ -96,7 +92,6 @@
       "cell_type": "code",
       "metadata": {
         "id": "AiJKd6uLE9Gr",
-        "colab_type": "code",
         "colab": {
           "base_uri": "https://localhost:8080/",
           "height": 468
@@ -106,7 +101,7 @@
       "source": [
         "!cd DenseDepth; python test.py"
       ],
-      "execution_count": 3,
+      "execution_count": null,
       "outputs": [
         {
           "output_type": "stream",
@@ -146,7 +141,6 @@
       "cell_type": "code",
       "metadata": {
         "id": "HjzqM74-FfyL",
-        "colab_type": "code",
         "colab": {
           "base_uri": "https://localhost:8080/",
           "height": 627
@@ -160,7 +154,7 @@
         "plt.figure(figsize=(20,20))\n",
         "plt.imshow( io.imread('./DenseDepth/test.png') )"
       ],
-      "execution_count": 7,
+      "execution_count": null,
       "outputs": [
         {
           "output_type": "execute_result",
@@ -191,20 +185,13 @@
     {
       "cell_type": "code",
       "metadata": {
-        "id": "Ra7GEtQBHUMS",
-        "colab_type": "code",
-        "colab": {}
+        "id": "Ra7GEtQBHUMS"
       },
       "source": [
-        "!cd DenseDepth/PyTorch; python test_pytorch.py\n",
-        "from matplotlib import pyplot as plt\n",
-        "from skimage import io\n",
-        "\n",
-        "plt.figure(figsize=(20,20))\n",
-        "plt.imshow( io.imread('./DenseDepth/test.png') )"
+        "!cd DenseDepth/PyTorch; python test_pytorch.py --cuda 1"
       ],
-      "execution_count": 0,
+      "execution_count": null,
       "outputs": []
     }
   ]
-}
+}

PyTorch/load_weight_from_keras.py → PyTorch/test_pytorch.py

@@ -19,13 +19,18 @@
 from torchvision import models
 import torch.nn.functional as F
 
-from pytorch_model import PTModel
+from model_pt import PTModel
 
-# Argument Parser
+device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+#device = 'cpu'
+#Argument Parser
 parser = argparse.ArgumentParser(description='High Quality Monocular Depth Estimation via Transfer Learning')
 parser.add_argument('--model', default='../nyu.h5', type=str, help='Trained Keras model file.')
 parser.add_argument('--input', default='../examples/*.png', type=str, help='Input filename or folder.')
+parser.add_argument('--cuda', default=1, type=int, help='Enable of Disbale Cuda')
 args = parser.parse_args()
+if args.cuda==0:
+  device = 'cpu'
 
 # Custom object needed for inference and training
 custom_objects = {'BilinearUpSampling2D': BilinearUpSampling2D, 'depth_loss_function': None}
@@ -84,7 +89,14 @@
 
 
 pytorch_model.load_state_dict(keras_state_dict)
+# pytorch_model = torch.load('depth_3.pth')
 pytorch_model.eval()
+#torch.save(pytorch_model,"depth_3.pth")
+pytorch_model.to(device)
+if device.__eq__('cuda'):
+  print("Loaded model to GPU")
+else: 
+  print("Loaded model to CPU")
 
 
 def my_DepthNorm(x, maxDepth):
@@ -94,20 +106,27 @@ def my_predict(model, images, minDepth=10, maxDepth=1000):
 
   with torch.no_grad():
     # Compute predictions
-    predictions = model(images)
+    predictions = model(images.to(device))
 
     # Put in expected range
-  return np.clip(my_DepthNorm(predictions.numpy(), maxDepth=maxDepth), minDepth, maxDepth) / maxDepth
-
+  return np.clip(my_DepthNorm(predictions.cpu().numpy(), maxDepth=maxDepth), minDepth, maxDepth) / maxDepth
+import time
 # # Input images
 inputs = load_images( glob.glob(args.input) ).astype('float32')
+
 pytorch_input = torch.from_numpy(inputs[0,:,:,:]).permute(2,0,1).unsqueeze(0)
-print(pytorch_input.shape)
+
+print("Input Shape = " + str(pytorch_input.shape))
 # print('\nLoaded ({0}) images of size {1}.'.format(inputs.shape[0], inputs.shape[1:]))
 
-# # Compute results
-output = my_predict(pytorch_model,pytorch_input[0,:,:,:].unsqueeze(0))
-print(output.shape)
+# Compute results (When it prdeicts on first it takes some time after that it runs fast you can check with using for loop)
+for i in range(10):
+  tic = time.time()
+  output = my_predict(pytorch_model,pytorch_input[0,:,:,:].unsqueeze(0))
+  toc  = time.time()
+  print("Time for test "+str(i)+" "+str(1000*(toc-tic))+" ms")
+
+print("Output Shape = " + str(output.shape))
 plt.imshow(output[0,0,:,:])
 plt.savefig('test.png')
 plt.show()

README.md

@@ -5,12 +5,16 @@
 
 Offical Keras (TensorFlow) implementaiton. If you have any questions or need more help with the code, contact the **first author**.
 
-**[Update]** Added a [Colab notebook](https://github.com/ialhashim/DenseDepth/blob/master/DenseDepth.ipynb) to try the method on the fly.
+**[Update]** Added a [Colab notebook](https://github.com/Avi241/DenseDepth/blob/master/DenseDepth.ipynb) to try the method on the fly.
 
 **[Update]** Experimental TensorFlow 2.0 implementation added.
 
 **[Update]** Experimental PyTorch code added.
 
+**[Update]** Implemented cuda version of Pytorch.
+
+**[Update]** Added Pytorch model for both Python3 and Python2. Please Find it [here.](https://drive.google.com/file/d/1Wi-w9uyErZwCn_x3DQeILntQgLdml72p/view?usp=share_link)
+
 ## Results
 
 * KITTI