cattaneod / pointnetvlad-pytorch Goto Github PK

Hi, thanks for your Pytorch implementation of PointNetVLAD.
I'm a new learner. When I read the code, I'm confused at PointNetVLAD:

According to my view:

• activation at line 59 represents the distribution of each point n belonging to the cluster k. (the weight)
• a at line 62 represents the learned cluster
• But what vlad at line 68 means? Where is the residuals and the sum of them according to points n?

Why not the following:

def forward(self, x):
        x = x.transpose(1, 3).contiguous()
        x = x.view((-1, self.max_samples, self.feature_size)) # [B,N,C]
        activation = torch.matmul(x, self.cluster_weights)
        if self.add_batch_norm:
            # activation = activation.transpose(1,2).contiguous()
            activation = activation.view(-1, self.cluster_size)
            activation = self.bn1(activation)
            activation = activation.view(-1,
                                         self.max_samples, self.cluster_size)
            # activation = activation.transpose(1,2).contiguous()
        else:
            activation = activation + self.cluster_biases
        activation = self.softmax(activation)
        activation = activation.view((-1, self.max_samples, self.cluster_size))
        
        a_sum = activation.sum(-2, keepdim=True)
        a = a_sum * self.cluster_weights2 # [B,C,k]
        
       ### ----------------- different-------------------
        N = x.shape[1]
        residual = x.unsqueeze(-1).repeat(1,1,1,self.cluster_size) - \
                          a.unsqueeze(1).repeat(1,N,1,1)  # [B,C,N,k]
        vlad = activation.unsqueeze(2) * residual
        vlad = torch.sum(vlad, dim=1) # [B,C,k]
       ### ----------------- different-------------------

        # intra-normalization and L2 normalize
        vlad = F.normalize(vlad, dim=1, p=2)    
        vlad = vlad.reshape((-1, self.cluster_size*self.feature_size))
        vlad = F.normalize(vlad, dim=1, p=2)

        # compress into a compact output
        vlad = torch.matmul(vlad, self.hidden1_weights)
        vlad = self.bn2(vlad)

        if self.gating:
            vlad = self.context_gating(vlad)

        return vlad

Is there anything wrong with my understanding?
Thanks in advance!

Hi,it is a great work, I want to do a test, can you provide a pre-trained model, thank you!

Thanks for your wonderful work, but i find in the file "train_pointnetvlad", the global variable "HARD_NEGATIVES={}" is never updated which means it is useless, is there something wrong here?

Why is batchsize=44 and how can I modify batchsize to be smaller? If you could answer me, I would greatly appreciate it

Thanks for your great work, the recall@1% on Oxford is higher than that in the paper, which is trained by the default config in your repository. Is it normal?

The import:
import config as cfg
is missing n the generate_training_tuples_baseline.py script, returning an error when defining the dataset directory name in line 10.

Hello, the default training file is oxford. How can I use the University section of the inhouse dataset for training? Thank you very much for your reply!

Hi, I'm a pytorch user and I'm really grateful that you open source this great work. However, I have some questions about the code. My question is, what are these two global variables HARD_NEGATIVES and TRAINING_LATENT_VECTORS in train_pointnetvlad.py? I notice that in the training process, before epoch 5, they won't be used since they are all empty. But after epoch 5 you will generate some latent_vectors via get_latent_vectors. So I wonder if they're something that help the model converge better, since according to its name (HARD_NAGATIVES), I guess you just select some harder negative samples from the training set. In other words, in the early stage of the training process, you just randomly select samples to train, and then after epoch 5 you select the most difficult samples to train the model more efficiently. Am I correct?

line 361, in process_data
ob[i] = torch.roll(ob[i], seed, dims=1)
RuntimeError: a view of a leaf Variable that requires grad is being used in an in-place operation.

Can you suggest why is this error is occurring?