Open zuixiaosanlang opened 2 years ago
I used abc data set as input, but I also got a bad mesh. Have you solved it, should i change other params in the code.
@zuixiaosanlang
@mabaorui
@yaping1119 sorry, i have not solved it. i give up when i get the bad mesh. if you solve it, please let me know, thanks
@zuixiaosanlang I tried shapenet data set and the effect was quite good, but the performance was poor on ABC. I think there should be something wrong with the setting
@yaping1119 Maybe the author will publish something later。
Hi, @mabaorui ,
I want to use your pretrain model in./pre_train_model/model-10201
, so I change OnSurfacePrior#775
to saver.restore(sess, a.out_dir + 'model-10201')
.
I ran command python onSurPrior.py --data_dir ./data/ --out_dir ./pre_train_model/ --CUDA 0 --INPUT_NUM 500 --epoch 30000 --input_ply_file input.ply --test
. But I got bad mesh,
I also try other .ply file, but the results are all the same.
And I change some paramater(line#425), but the results are bad.
How could I got the expected results? for every .ply file train a new model?
@zuixiaosanlang I tried shapenet data set and the effect was quite good, but the performance was poor on ABC. I think there should be something wrong with the setting
Hi, @yaping1119 Could you please share some shapenet file? ( this programe can generate the expected results ) Thanks in advanced.
@Bailey-24 ,The author uses /data/input.ply as an example of shapenet。
@yaping1119 , But the result show as above picture. ( The .off file is a curved surface)
@yaping1119 , But the result show as above picture. ( The .off file is a curved surface)
After training a new model, the result is pretty nice. Thanks.
I use the pretrained model, and my result is just like this, how could i get a better result, please.
i use my data input, but get a bad mesh, should i change other params in the code ? --data_dir ./data/ --out_dir ./train_net/ --CUDA 0 --INPUT_NUM 71642 --epoch 30000 --input_ply_file laser_71642.ply --train