Comments (4)
Hi, the evaluation script in the train folder is
python -m torch.distributed.launch --nproc_per_node=1 train.py --checkpoint ./pretrained_checkpoint/sam_vit_l_0b3195.pth --model-type vit_l --output work_dirs/hq_sam_l --eval --restore-model work_dirs/hq_sam_l/epoch_11.pth
In this script, we load sam_vit_l_0b3195.pth for the encoder output and load the additional parameters trained by ours with the argument --restore-model
. In training, we only learn a small number of parameters and save them. For evaluation on HQ dataset, we only need to load this group of parameters.
An example checkpoint for --restore-model
can be found here. You can also train it yourself.
from sam-hq.
from sam-hq.
Hi, when evaluating the four HQ datasets, we only need to load the mask_decoder of hq_sam. We do it this way because it saves storage space during training. For example,
--restore-model work_dirs/hq_sam_l/epoch_11.pth
Here epoch_11.pth
is the decoder part of the sam_hq_vit_l.pth
. An example checkpoint for --restore-model
can be found in this link. You can also train it yourself.
from sam-hq.
I faced this error too. I used the command python3 -m torch.distributed.launch --nproc_per_node=1 train.py --checkpoint ./pretrained_checkpoint/sam_vit_l_0b3195.pth --model-type vit_l --output work_dirs/hq_sam_l/ --eval --restore-model work_dirs/hq_sam_l/sam_hq_epoch_11.pth
where sam_hq_epoch_11.pth
was the final checkpoint that the training had saved. However when I tried with any other previous checkpoints, it worked as expected.
from sam-hq.
Related Issues (20)
- Data Loader throwing FileNotFound error After few epochs of training HOT 1
- About custom dataset. HOT 2
- Some issues in the fine-tuning process HOT 3
- I wanted to train the model with my images HOT 3
- How do I run the training script if I only have one graphics card? HOT 6
- Can the weight format generated by your training code be directly used in GroundedSAM HOT 2
- Hello author, how do you measure loss_mask and loss dice, do you have any references? HOT 2
- When I use ''multimask_output=True'', only one mask is generated HOT 1
- Killing Subprocess during training HOT 2
- Why does my implementation of sam-hq light takes longer time? HOT 1
- When using my own data for training, the first epoch reached an IOU value of over 0.9. May I ask if there is a problem? The data is 5000 HOT 1
- Where is the training script for Light HQ-SAM? HOT 1
- DecompressionBombWarning HOT 1
- How to export the tensorrt model? HOT 1
- Performing batch image detection using a for loop can lead to memory leakage. HOT 2
- Error when trying to load embedding image in browser using onnx HOT 3
- How to generate embedding file using SAMHQ? HOT 2
- how to generate a sliding comparison gif file in the Visual comparison part HOT 2
- The fine-tuned HQ-SAM model shows a significant improvement in accuracy. However, the val_iou_0 accuracy of HQ-SAM during training is very low. HOT 2
- how to train the model from previous saved checkpoints HOT 2
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 sam-hq.