SuLvXiangXin / zipnerf-pytorch

Unofficial implementation of ZipNeRF
Apache License 2.0
783 stars 85 forks source link

Floating artifacts with custom dataset #60

Open shairatabassum opened 1 year ago

shairatabassum commented 1 year ago

Hi, first of all, thank you for the implementation.

I have trained and rendered the model with some existing scenes from the 360 dataset such as bicycle, kitchen, room. The rendering is almost-perfect for each case. However, when it comes to custom datasets, I'm having a lot of ghostly artifacts. Some of the outputs are available here.

Could you please take a look and provide any possible reasons or suggestions? Thank you!

I'm using the following commands for training and rendering:

python train.py \
    --gin_configs=configs/360.gin \
    --gin_bindings="Config.data_dir = '${DATA_DIR}'" \
    --gin_bindings="Config.exp_name = '${EXP_NAME}'" \
    --gin_bindings="Config.factor = 4" \
    --gin_bindings="Config.batch_size = 8192"

python render.py \
    --gin_configs=configs/360.gin \
    --gin_bindings="Config.data_dir = '${DATA_DIR}'" \
    --gin_bindings="Config.exp_name = '${EXP_NAME}'" \
    --gin_bindings="Config.render_path = True" \
    --gin_bindings="Config.render_path_frames = 480" \
    --gin_bindings="Config.render_video_fps = 120” \
    --gin_bindings="Config.factor = 4"
SuLvXiangXin commented 1 year ago

Maybe your dataset is taken in a varying light condition? Try using 360_glo.gin to train it again.

alex89607 commented 11 months ago

Hi! Thanks for implementation! The same fog. and when the use using 360_glo.gin too.

hg111 commented 8 months ago

@alex89607 , @shairatabassum @SuLvXiangXin I was wondering, did you find a solution to the "fog"? I am getting the same especially when the camera get's too close to the object