ImageDecoder libtiff implementation #1264
Conversation
Signed-off-by: Joaquin Anton <janton@nvidia.com>
Signed-off-by: Joaquin Anton <janton@nvidia.com>
Signed-off-by: Joaquin Anton <janton@nvidia.com>
Signed-off-by: Joaquin Anton <janton@nvidia.com>
jantonguirao
force-pushed
the
libtiff_impl
branch
from
a7b1c31
to
a8e2b80
Compare
Signed-off-by: Joaquin Anton <janton@nvidia.com>
Signed-off-by: Joaquin Anton <janton@nvidia.com>
Signed-off-by: Joaquin Anton <janton@nvidia.com>
jantonguirao
changed the title
|
CI MESSAGE: [907590]: BUILD STARTED |
Signed-off-by: Joaquin Anton <janton@nvidia.com>
jantonguirao
force-pushed
the
libtiff_impl
branch
from
1fafb69
to
83f1475
Compare
|
CI MESSAGE: [907594]: BUILD STARTED |
Signed-off-by: Joaquin Anton <janton@nvidia.com>
|
CI MESSAGE: [907594]: BUILD FAILED |
|
!build |
|
CI MESSAGE: [907791]: BUILD STARTED |
|
CI MESSAGE: [907791]: BUILD FAILED |
Signed-off-by: Joaquin Anton <janton@nvidia.com>
There was a problem hiding this comment.
That's incorrect design here.
IMHO You should leverage the object-oriented design and do it like:
image_factory.cc
std::unique_ptr<Image>
ImageFactory::CreateImage(const uint8_t *encoded_image, size_t length, DALIImageType image_type) {
[...]
} else if (CheckIsTiff(encoded_image, length)) {
if (DALI_USE_LIBTIFF)
return std::make_unique<TiffImage_LibTiff>(encoded_image, length, image_type);
return std::make_unique<TiffImage>(encoded_image, length, image_type);
}
[...]
}
And create class TiffImage_LibTiff : public TiffImage. That way, PeekDims is the same for both implementation (which I think it should be), fallback for OpenCV is guaranteed and you only write DecodeImpl function.
|
CI MESSAGE: [907825]: BUILD STARTED |
|
CI MESSAGE: [907825]: BUILD FAILED |
Signed-off-by: Joaquin Anton <janton@nvidia.com>
I agree. Fixed |
jantonguirao
force-pushed
the
libtiff_impl
branch
from
5bc4017
to
4e94ec5
Compare
|
CI MESSAGE: [908043]: BUILD STARTED |
Signed-off-by: Joaquin Anton <janton@nvidia.com>
jantonguirao
force-pushed
the
libtiff_impl
branch
from
ec8e037
to
6bdc239
Compare
|
CI MESSAGE: [908055]: BUILD STARTED |
|
CI MESSAGE: [908089]: BUILD STARTED |
|
CI MESSAGE: [908055]: BUILD FAILED |
|
CI MESSAGE: [908110]: BUILD STARTED |
|
CI MESSAGE: [908089]: BUILD FAILED |
|
CI MESSAGE: [908110]: BUILD FAILED |
Signed-off-by: Joaquin Anton <janton@nvidia.com>
|
CI MESSAGE: [909512]: BUILD STARTED |
|
CI MESSAGE: [909527]: BUILD STARTED |
|
CI MESSAGE: [909536]: BUILD STARTED |
|
|
||
| const auto ifd_offset = buffer.Read<uint32_t>(4); | ||
| const auto entry_count = buffer.Read<uint16_t>(ifd_offset); | ||
| bool width_read = false, height_read = false; | ||
| size_t width, height; | ||
| size_t width = 0, height = 0; |
There was a problem hiding this comment.
Just change the type here to int64_t and avoid the cast in return?
There was a problem hiding this comment.
Will take care of it in #1280
|
CI MESSAGE: [909527]: BUILD FAILED |
|
|
||
| #define DALI_ERROR(str) \ |
There was a problem hiding this comment.
Either align the backslashes
if (a) \
asdf \
or move them to one space after end of line
if (a) \
asdf \
There was a problem hiding this comment.
Will do in #1280
|
CI MESSAGE: [909536]: BUILD FAILED |
|
CI MESSAGE: [909602]: BUILD STARTED |
|
CI MESSAGE: [909602]: BUILD PASSED |
|
CI MESSAGE: [909708]: BUILD STARTED |
|
CI MESSAGE: [909708]: BUILD FAILED |
|
CI MESSAGE: [909602]: BUILD PASSED |
Signed-off-by: Joaquin Anton <janton@nvidia.com> Signed-off-by: Jianjun Liu <00liujj@163.com>
Signed-off-by: Joaquin Anton janton@nvidia.com
Why we need this PR?
What happened in this PR?
JIRA TASK: [DALI-1038]