ace19-dev / mvcnn-tf

Basic reference for Multi View Classification - mvcnn

MVCNN

Data

• Download modelnet10-Class Orientation-aligned Subset
  • You can create 2D dataset from 3D objects (.obj, .stl, and .off), using BlenderPhong
• Directly you get dataset https://drive.google.com/file/d/0B4v2jR3WsindMUE3N2xiLVpyLW8/view from https://github.com/RBirkeland/MVCNN-PyTorch/blob/master/README.md

Quick Start

• prepare multi-view image
• train.py

Evaluation

• eval.py

Retrieval

• deep metric learning
• extract features by train.py
• prepare query/gallery sources
• retrieve.py

TODO

• improve retrieval module.
• balanced sampling batch.

Notice

• The network architecture used in our experiments is relatively shallow to allow for fast training and inference.

References from

• https://github.com/WeiTang114/MVCNN-TensorFlow
• https://github.com/WeiTang114/BlenderPhong
• https://github.com/RBirkeland/MVCNN-PyTorch/blob/master/README.md
• https://github.com/tensorflow/models/tree/master/official/resnet
• https://arxiv.org/abs/1812.00442
• https://github.com/ace19-dev/gvcnn

mvcnn-tf's Issues

How to draw precision-recall curves?

Hello,thank you for your work,How to draw precision-recall curves?Is there code for the diagram?

version problems

when i am trying run the code,i faced some problems:
Traceback (most recent call last):
File "D:/Project/mvcnn-tf/train.py", line 9, in
from utils import train_utils
File "D:\Project\mvcnn-tf\utils\train_utils.py", line 19, in
slim = tf.contrib.slim
AttributeError: module 'tensorflow' has no attribute 'contrib'

my tensorflow version is 2.0
when my version changed 1.12.0,i faced problems like:
Traceback (most recent call last):
File "D:/Project/mvcnn-tf/train.py", line 352, in
if tf.io.gfile.exists(FLAGS.train_logdir):#判断目录是否存在
AttributeError: module 'tensorflow._api.v1.io' has no attribute 'gfile'

i don't know how to solve,thank you

About code content

Thank you so much for sharing, but I have a problem, could you please help me solve it?
In train.py, there is " training_tf_filenames = os.path.join(FLAGS.dataset_dir, 'train.record')" what does train.record means? The name of folder? Or a list contains all the names of data inside this path?

tf.compat.v1 has no attribute'v1'

Thanks your work,I met this problem:
reuse=tf.compat.v1.AUTO_REUSE,
AttributeError: 'module' object has no attribute 'v1'
How I solve this problem?