volotat / diffmorph Goto Github PK

Hello,
I'm trying to use DiffMorph with google colab (because I'm not able to understand how to make it work otherwise.. (if you guys btw can help, I'll be glad!).
However, maybe I'm missing something, but I'm not able to make work the step 2 "Step 2: Add images to the "DiffMorph/input" directory." ---> It seems that there is no code to run, and I don't know where is that directory, where to find it etc.

So I need some help please here..
Thanks a lot!

Can this program run on Macbook with M1 chip? I been having a lot of problems with the versions of tensorflow and numpy.

Hi,

Would the learned warp maps be able to translate a given pixel location in the source image to a location in the target image?

Hello, I was trying to execut the morph.py command, however I got an error:

"(tf) C:\Users\chris\Desktop\DiffMorph-master\DiffMorph-master>morph.py -s images/test(2).jpg -t images/test(1).jpg
Traceback (most recent call last):
File "C:\Users\chris\Desktop\DiffMorph-master\DiffMorph-master\morph.py", line 3, in
import numpy as np
ModuleNotFoundError: No module named 'numpy'"

I think I'll need some help here please ^^
Thanks a lot!

Hi developers,

Thanks for your codes.
I ran into a problem for few days.
Due to the limitations of my research zone, all dependencies can only download from anaconda and then ship to my conda environment. There exists numerous problems during env construction.

Would you mind providing the docker image so that we can build the env properly.
I will appreciate your help with this situation.

Hello.
I needed to install the CUDA 11.3 library to fix the following error message for my RTX 3080ti
"Could not load dynamic library 'cudart64_110.dll'"

It's on the tensorflow page that you need NVidia libraries https://www.tensorflow.org/install/pip#windows-native
but this may not be obvious to those that are installing it for the first time just via pip.

Thanks.

Hello. Just made your code work on my computer. Tensorflow isn't the only dependency requiered, you may want to update your readme. You also need tensorflow-add_ons and cv2.

They can all be install through this command

pip3 install tensorflow tensorflow_addons opencv-python

This would make your code more accesible.
Cheers!

Hi
Executing just morph.py -s images/xxx.jpg -t images/xxx.jpg [-e 1000 -a 0.8 -m 0.8 -w 0.3] gives me just "command not found. "Executing : ./morph.py -s images/trump.jpg -t images/pence.jpg [-e 1000 -a 0.8 -m 0.8 -w 0.3] gives me this:
./morph.py: line 1: import: command not found
./morph.py: line 2: import: command not found
./morph.py: line 3: import: command not found
./morph.py: line 4: import: command not found
./morph.py: line 5: import: command not found
./morph.py: line 8: TRAIN_EPOCHS: command not found
./morph.py: line 10: im_sz: command not found
./morph.py: line 11: mp_sz: command not found
./morph.py: line 13: warp_scale: command not found
./morph.py: line 14: mult_scale: command not found
./morph.py: line 15: add_scale: command not found
./morph.py: line 16: add_first: command not found
./morph.py: line 19: @tf.function: command not found
./morph.py: line 20: syntax error near unexpected token (' ./morph.py: line 20: def warp(origins, targets, preds_org, preds_trg):'
So it doesent work for me in Ubuntu 20.04 with tensorflow 2.4.0

One step further i tried :
DiffMorph$ python3 morph.py -s images/xxx.jpg -t images/xxx.jpg [-e 1000 -a 0.8 -m 0.8 -w 0.3]
2021-01-09 16:47:46.888345: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'libcudart.so.11.0'; dlerror: libcudart.so.11.0: cannot open shared object file: No such file or directory
2021-01-09 16:47:46.888361: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
Traceback (most recent call last):
File "morph.py", line 4, in
import cv2
ModuleNotFoundError: No module named 'cv2'

Have no idea whats wrong here !

While running use_wrap_maps method, the script uses memory proportional to the number of steps. Making it difficult to generate higher resolution images and higher number of frames.

transition between images is very gray

I use -e 10000 -a 0.4 -m 0.4 -w 0.1

Latest try:

python morph.py -s images/img_1.jpg -t images/img_2.jpg

2021-01-12 17:23:25.848313: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN)to use the following CPU instructions in performance-critical operations: SSE4.1 SSE4.2 AVX AVX2 FMA

To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.

2021-01-12 17:23:25.869795: I tensorflow/core/platform/profile_utils/cpu_utils.cc:104] CPU Frequency: 3600000000 Hz

2021-01-12 17:23:25.870428: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x557665775f10 initialized for platform Host (this does not guarantee that XLA will be used). Devices:

2021-01-12 17:23:25.870473: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version

Epoch 100, Loss: 0.07585608214139938

Epoch 200, Loss: 0.056177154183387756

Epoch 300, Loss: 0.047623421996831894

Epoch 400, Loss: 0.04261484742164612

Epoch 500, Loss: 0.03922488912940025

Epoch 600, Loss: 0.036724839359521866

Epoch 700, Loss: 0.03478630632162094

Epoch 800, Loss: 0.03322023153305054

Epoch 900, Loss: 0.031917549669742584

Epoch 1000, Loss: 0.030810609459877014

Image #10 produced.

Image #20 produced.

Image #30 produced.

Image #40 produced.

Image #50 produced.

Image #60 produced.

Image #70 produced.

Image #80 produced.

Image #90 produced.

Image #100 produced.

Result video saved.

So far it looks promising but in the morph folder the video is just 630 kb and black when trying to play. 2 pictures maps.jpg +result.jpg look better see https://i.imgur.com/BbhNnsZ.jpg and https://i.imgur.com/6pYS0gw.jpg . Seems like i am coming closer but have no idea why the video is not created

Which parameters were used to generate car morph from example? How many epochs is needed to generate that result?

I'm newbie in neural networks so sorry for possible stupid question

Issue Description

After setting up the virtualenv, the following error was printed to terminal when running python3 morph.py -s images/image_1.png -t images/image_2.png

Traceback (most recent call last):
  File "./morph.py", line 8, in <module>
    from tqdm import tqdm
ModuleNotFoundError: No module named 'tqdm'

Environment

• Ubuntu 20.04 Focal Fossa
• Nvidia Geforce GTX 1060

Steps to Reproduce

Follow the steps below to reproduce the error described above:

virtualenv -p python3 venv
source venv/bin/activate
pip3 install -r requirements.txt
python3 morph.py -s images/image_1.png -t images/image_2.png

Others

This issue has been resolved. See below for the solution. Posting this issue to simply publish the solution.

Running the colab project fails at a number of points, starting with this error:

ERROR: Could not find a version that satisfies the requirement numpy==1.23.1

Unfortunately don't have enough time to troubleshoot this one myself - Anyone else experience this or know a fix, or know what I'm doing wrong? - I am very new to colaboratory