manoja328 / rodeo Goto Github PK

View Code? Open in Web Editor NEW

26.0 26.0 2.0 131 KB

Official implementation of "RODEO: Replay for Online Object Detection", BMVC 2020

Home Page: https://www.manojacharya.com

License: MIT License

Python 99.52% Shell 0.48%

computer-vision continual-learning lifelong-learning object-detection pytorch streaming

rodeo's People

Contributors

Stargazers

Watchers

Forkers

vishwa30 noodlesz

rodeo's Issues

No such file: 'datasets/edboxes_voc07_2000_new.pkl'

Hi, where can I find the required file when running train_better.py?

[Questions] About the Edge Box Proposals Extraction

Hi Manoj,

Thank you for your great work! I notice that the README only provides the Edgebox proposals file for the VOC 2007, and I also check the file extract_coco_features.py that there do not include any EdgeBox proposals for the MSCOCO 2014.

This makes me confused because based on your paper, you mentioned that you are using the edge box proposals following the ILWFOD [CVPR 2017] in Section 5.5 Implementation Details. Do I misunderstand this detail? Is there any reason for the missing edge box proposals for the MSCOCO 2014 dataset?

The other question is: how can I obtain the edge box, e.g. your provided edge box proposals file for VOC 2007?

Best,
Mingfu

Missing the parameter pq_features when I run the train_better.py file

Hi, an error occurs when I run the train_better.py file:

Traceback (most recent call last):
  File "train_better.py", line 296, in <module>
    loss_dict = model(images, targets)
  File "/home/cy/.conda/envs/RODEO/lib/python3.7/site-packages/torch/nn/modules/module.py", line 532, in __call__
    result = self.forward(*input, **kwargs)
  File "/data/cy/rodeo-master/frcnn_mod.py", line 57, in forward
    raise ValueError("In training mode, targets should be passed")
ValueError: In training mode, targets should be passed

I find I need a parameter pq_features here:

rodeo/frcnn_mod.py

Lines 43 to 77 in c7f340a

    
               def forward(self, images, pq_features, targets=None): 
        
                   """ 
        
                   Arguments: 
        
                       images (list[Tensor]): images to be processed 
        
                       targets (list[Dict[Tensor]]): ground-truth boxes present in the image (optional) 
        
                   Returns: 
        
                       result (list[BoxList] or dict[Tensor]): the output from the model. 
        
                           During training, it returns a dict[Tensor] which contains the losses. 
        
                           During testing, it returns list[BoxList] contains additional fields 
        
                           like `scores`, `labels` and `mask` (for Mask R-CNN models). 
        
                   """ 
        
                   if self.training and targets is None: 
        
                       raise ValueError("In training mode, targets should be passed") 
        
                   original_image_sizes = [img.shape[-2:] for img in images] 
        
                   images, targets = self.transform(images, targets) 
        
                   features = self.backbone(pq_features) 
        
                   if isinstance(features, torch.Tensor): 
        
                       features = OrderedDict([(0, features)]) 
        
                   proposals, proposal_losses = self.rpn(images, features, targets) 
        
                   proposals = [p.to(features[0].device) for p in proposals] 
        
                   detections, detector_losses = self.roi_heads(features, proposals, images.image_sizes, targets) 
        
                   detections = self.transform.postprocess(detections, images.image_sizes, original_image_sizes) 
        
                   losses = {} 
        
                   losses.update(detector_losses) 
        
                   losses.update(proposal_losses) 
        
                   if self.training: 
        
                       return losses 
        
                   return detections

but in train_better.py file, there are only images and targets:

rodeo/train_better.py

Line 294 in c7f340a

loss_dict = model(images, targets)

Is the problem here? How to solve it?
Looking for your reply. Thanks.

This is the worst code i had ever seen

Could not image who did this

The incremental process

Hello,I have some doubts about the incremental process. Can you elaborate on how program files are incremental? For example, what should I do after I initialize G with the VOC dataset?

Where is get_data128 function?

So pleasure to view this project.
I have a question about get_distillationinfo in train_better.py

def get_distillinfo(model, dl):
    save = {}
    print("dumping info ......")
    model.eval()
    with torch.no_grad():
        for ii, (images, targets) in tqdm(enumerate(dl), total=len(dl)):
            images = list(image.to(device) for image in images)
            targets = [{k: v.to(device) for k, v in t.items()} for t in targets]
            for image, target in zip(images, targets):
                image_id = '{0:06d}'.format(target['image_id'].item())
                info = model.get_data128([image], [target])
                save[image_id] = info
        return save

where can i find the get_data128() function？ I don't see it in frcnn_mod.py ?

Not able to create environment

Hello, I am getting Solving environment failed error when trying to create environment using the given yml file. This can be mostly because I am using Windows machine and there may be some platform dependencies issues. So is there a way I can run this in Windows as well?

manoja328 / rodeo Goto Github PK

rodeo's People

Contributors

Stargazers

Watchers

Forkers

rodeo's Issues

No such file: 'datasets/edboxes_voc07_2000_new.pkl'

[Questions] About the Edge Box Proposals Extraction

Missing the parameter pq_features when I run the train_better.py file

This is the worst code i had ever seen

The incremental process

Where is get_data128 function?

Not able to create environment

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

	def forward(self, images, pq_features, targets=None):
	"""
	Arguments:
	images (list[Tensor]): images to be processed
	targets (list[Dict[Tensor]]): ground-truth boxes present in the image (optional)

	Returns:
	result (list[BoxList] or dict[Tensor]): the output from the model.
	During training, it returns a dict[Tensor] which contains the losses.
	During testing, it returns list[BoxList] contains additional fields
	like `scores`, `labels` and `mask` (for Mask R-CNN models).

	"""
	if self.training and targets is None:
	raise ValueError("In training mode, targets should be passed")
	original_image_sizes = [img.shape[-2:] for img in images]
	images, targets = self.transform(images, targets)
	features = self.backbone(pq_features)

	if isinstance(features, torch.Tensor):
	features = OrderedDict([(0, features)])
	proposals, proposal_losses = self.rpn(images, features, targets)
	proposals = [p.to(features[0].device) for p in proposals]
	detections, detector_losses = self.roi_heads(features, proposals, images.image_sizes, targets)
	detections = self.transform.postprocess(detections, images.image_sizes, original_image_sizes)

	losses = {}
	losses.update(detector_losses)
	losses.update(proposal_losses)


	if self.training:
	return losses

	return detections