Comments (7)
@FishYuLi Some other known differences are:
- The detectron changes the learning rate by 2 times, there has 0.2~0.4 mAP improvement.
- The implementation of weight-decay in MXNET and Caffe2 has a little difference, this may lead 0.1~0.3 mAP gaps compared with the best results, depending on how many GPUs do you use.
- We closed the warmup.
- The detectron adopts Xariver init, we use random Gaussian.
- We use softnms=0.6 and the detectron use NMS=0.5. For plain ROI, the softnms will have 0.4~0.5 mAP gain than NMS. But for ROI-Align, there is only 0.2 mAP gain.
from relation-networks-for-object-detection.
A difference is that detectron use RoIAlign, where we use RoIPool.
from relation-networks-for-object-detection.
@chengdazhi Yes, I notice that. But I'm not sure that if RoIAlign can make a difference of nearly 1.7 points. The paper of Mask RCNN shows that this may lead to an improvement of about 1 point, which makes me a little confused.
Thanks a lot. I may try RoIAlign later.
from relation-networks-for-object-detection.
Another possibility is that we train FPN in a two stage manner, this speeds up training, but can harm accuracy.
from relation-networks-for-object-detection.
@chengdazhi 38.5 is also the result of a two-stage manner with pre-computed proposals and 1x schedule. They got 39.4 with end2end training.
from relation-networks-for-object-detection.
different pretrained resnet models could be another possible cause
from relation-networks-for-object-detection.
@stupidZZ Sounds more reasonable. Thanks! : )
from relation-networks-for-object-detection.
Related Issues (20)
- learn nms does not work good on pascal voc HOT 4
- errors when calling output_shapes HOT 2
- why adding nms_embedding_feat and nms_attention_1 together
- have a question:what the "non_gt_index" means in this project??thanks! HOT 2
- How to visualize the bounding box of a new image using demo model? HOT 1
- NotImplementedError HOT 1
- There are many overlap bboxs in the test result, does it effect the mAP?
- Question about "nongt_dim" HOT 1
- What does mx.sym.full stand for? HOT 1
- Continue training from specific epoch model HOT 1
- How to generate "previously proposals" for another dataset beyond COCO to train with RPN HOT 5
- rois -> sliced_rois, why "sliced_rois" start from 1?
- Question about "nongt_dim" HOT 4
- does it works based on vgg16?
- windows ? or ubuntu
- mxnet.base.MXNetError: Error in operator _plus2: [06:13:23] src/operator/contrib/./../elemwise_op_common.h:135: Check failed: assign(&dattr, vec.at(i)): Incompatible attr in node _plus2 at 1-th input: expected [313,16,300], got [19,16,18]
- Validation
- Compile MXNet HOT 1
- Where is the "utils" dictory?
- relation code HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from relation-networks-for-object-detection.