search

rowanz / neural-motifs

Code for Neural Motifs: Scene Graph Parsing with Global Context (CVPR 2018)
https://rowanzellers.com/neuralmotifs
MIT License
525 stars 115 forks source link

Not sure RelModel is learning #43

Closed galsk87 closed 6 years ago

galsk87 commented 6 years ago

while running train_rels.py -m sgcls -model motifnet -order leftright -nl_obj 2 -nl_edge 4 -b 6 -clip 5 \ -p 100 -hidden_dim 512 -pooling_dim 4096 -lr 1e-3 -ngpu 1 -ckpt checkpoints/vgdet/vg-24.tar \ -save_dir checkpoints/motifnet2 -nepoch 50 -use_bias

my rel loss is constantly around 0.15-0.12 is this the loss you got while training , did you see any improvements, do you think the model might not learn like expected?

i'm getting similar results to what you published.

rowanz commented 6 years ago

if you're getting good results, then the model is learning, right? One thing to note is that there often isn't much to learn on top of the frequency baseline 😄

galsk87 commented 6 years ago

You are not inputting it into the motifnet model though right?

On Wed, Nov 28, 2018, 9:23 PM Rowan Zellers <notifications@github.com wrote:

if you're getting good results, then the model is learning, right? One thing to note is that there often isn't much to learn on top of the frequency baseline 😄

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/rowanz/neural-motifs/issues/43#issuecomment-442572198, or mute the thread https://github.com/notifications/unsubscribe-auth/AbMDMgWn3LqT9Agqb57FLrIsnWYvJdQLks5uzuK_gaJpZM4YxW_y .