* @file tensordec.c
* @date 26 Mar 2018
* @brief GStreamer plugin to convert tensors (as a filter for other general neural network filters) to other media types
- * @see http://github.com/TO-BE-DETERMINED-SOON
+ * @see https://github.com/nnsuite/nnstreamer
* @see https://github.sec.samsung.net/STAR/nnstreamer
* @author Jijoong Moon <jijoong.moon@samsung.com>
* @bug No known bugs except for NYI items
#endif
#include <string.h>
-#include <gst/gst.h>
-#include <glib.h>
-
#include "tensordec.h"
+/**
+ * @brief Macro for debug mode.
+ */
+#ifndef DBG
+#define DBG (!self->silent)
+#endif
+
+#define silent_debug(...) \
+ debug_print (DBG, __VA_ARGS__)
+
+#define silent_debug_caps(caps,msg) if (DBG) { \
+ if (caps) { \
+ GstStructure *caps_s; \
+ gchar *caps_s_string; \
+ guint caps_size, caps_idx; \
+ caps_size = gst_caps_get_size (caps);\
+ for (caps_idx = 0; caps_idx < caps_size; caps_idx++) { \
+ caps_s = gst_caps_get_structure (caps, caps_idx); \
+ caps_s_string = gst_structure_to_string (caps_s); \
+ debug_print (TRUE, msg " = %s\n", caps_s_string); \
+ g_free (caps_s_string); \
+ } \
+ } \
+}
+
GST_DEBUG_CATEGORY_STATIC (gst_tensordec_debug);
#define GST_CAT_DEFAULT gst_tensordec_debug
-/* Filter signals and args */
-enum
-{
- /* FILL ME */
- LAST_SIGNAL
-};
-
+/** properties */
enum
{
PROP_0,
};
/**
+ * @brief Flag to print minimized log.
+ */
+#define DEFAULT_SILENT TRUE
+
+/**
* @brief The capabilities of the inputs
- *
- * In v0.0.1, this is 3-d tensor, [color][height][width]
- *
*/
static GstStaticPadTemplate sink_factory = GST_STATIC_PAD_TEMPLATE ("sink",
GST_PAD_SINK,
GST_STATIC_CAPS (GST_TENSOR_CAP_DEFAULT)
);
-/* the capabilities of the outputs
- *
- * In v0.0.1, this is "bitmap" image stream
+/**
+ * @brief The capabilities of the outputs
*/
static GstStaticPadTemplate src_factory = GST_STATIC_PAD_TEMPLATE ("src",
GST_PAD_SRC,
#define gst_tensordec_parent_class parent_class
G_DEFINE_TYPE (GstTensorDec, gst_tensordec, GST_TYPE_BASE_TRANSFORM);
+/** GObject vmethod implementations */
static void gst_tensordec_set_property (GObject * object, guint prop_id,
const GValue * value, GParamSpec * pspec);
static void gst_tensordec_get_property (GObject * object, guint prop_id,
GValue * value, GParamSpec * pspec);
-/* GstBaseTransformer vmethod implementations */
+/** GstBaseTransform vmethod implementations */
static GstFlowReturn gst_tensordec_transform (GstBaseTransform * trans,
GstBuffer * inbuf, GstBuffer * outbuf);
static GstFlowReturn gst_tensordec_transform_ip (GstBaseTransform * trans,
GstPadDirection direction, GstCaps * caps, GstCaps * othercaps);
static gboolean gst_tensordec_set_caps (GstBaseTransform * trans,
GstCaps * incaps, GstCaps * outcaps);
-/* GObject vmethod implementations */
+static gboolean gst_tensordec_transform_size (GstBaseTransform * trans,
+ GstPadDirection direction, GstCaps * caps, gsize size,
+ GstCaps * othercaps, gsize * othersize);
+
+/**
+ * @brief Parse structure and return media caps
+ * @param self "this" pointer
+ * @param structure structure to be interpreted
+ */
+static GstCaps *
+gst_tensordec_media_caps_from_structure (GstTensorDec * self,
+ const GstStructure * structure)
+{
+ GstTensorConfig config;
+ GstCaps *result = NULL;
+
+ if (gst_tensor_config_from_structure (&config, structure)) {
+ result = gst_tensor_media_caps_from_config (&config);
+ }
+
+ if (result == NULL) {
+ /** raw caps for video, audio */
+ GstStaticCaps caps_video = GST_STATIC_CAPS (GST_TENSOR_VIDEO_CAPS_STR);
+ GstStaticCaps caps_audio = GST_STATIC_CAPS (GST_TENSOR_AUDIO_CAPS_STR);
+
+ result = gst_caps_merge (gst_static_caps_get (&caps_video),
+ gst_static_caps_get (&caps_audio));
+ result = gst_caps_simplify (result);
+ }
+
+ return result;
+}
+
+/**
+ * @brief Check tensor config is consistent
+ * @param self "this" pointer to check consistency
+ * @param t_info newly configured tensor metadata
+ */
+static gboolean
+gst_tensordec_check_consistency (GstTensorDec * self, GstTensorConfig * config)
+{
+ g_return_val_if_fail (self != NULL, FALSE);
+ g_return_val_if_fail (config != NULL, FALSE);
+
+ if (self->configured) {
+ return gst_tensor_config_is_same (&self->tensor_config, config);
+ }
+
+ /** not configured yet */
+ return FALSE;
+}
/**
* @brief initialize the tensordec's class
*/
static void
-gst_tensordec_class_init (GstTensorDecClass * g_class)
+gst_tensordec_class_init (GstTensorDecClass * klass)
{
GObjectClass *gobject_class;
GstElementClass *gstelement_class;
GstBaseTransformClass *trans_class;
- GstTensorDecClass *klass;
- klass = (GstTensorDecClass *) g_class;
trans_class = (GstBaseTransformClass *) klass;
gstelement_class = (GstElementClass *) trans_class;
gobject_class = (GObjectClass *) gstelement_class;
g_object_class_install_property (gobject_class, PROP_SILENT,
g_param_spec_boolean ("silent", "Silent", "Produce verbose output ?",
- FALSE, G_PARAM_READWRITE));
+ DEFAULT_SILENT, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
gst_element_class_set_details_simple (gstelement_class,
"TensorDec",
- "Convert tensor stream to media stream ",
+ "Convert tensor stream to media stream",
"Converts tensor stream of C-Array for neural network framework filters to audio or video stream",
"Jijoong Moon <jijoong.moon@samsung.com>");
gst_element_class_add_pad_template (gstelement_class,
gst_static_pad_template_get (&src_factory));
- /* Refer: https://gstreamer.freedesktop.org/documentation/design/element-transform.html */
+ /** Refer: https://gstreamer.freedesktop.org/documentation/design/element-transform.html */
trans_class->passthrough_on_same_caps = FALSE;
- /* Processing units */
+ /** Processing units */
trans_class->transform = GST_DEBUG_FUNCPTR (gst_tensordec_transform);
trans_class->transform_ip = GST_DEBUG_FUNCPTR (gst_tensordec_transform_ip);
- /* Negotiation units */
+ /** Negotiation units */
trans_class->transform_caps =
GST_DEBUG_FUNCPTR (gst_tensordec_transform_caps);
trans_class->fixate_caps = GST_DEBUG_FUNCPTR (gst_tensordec_fixate_caps);
trans_class->set_caps = GST_DEBUG_FUNCPTR (gst_tensordec_set_caps);
- /** Allocation units
- * @TODO Need to add Allocation units depending on the tensor size.
- */
+ /** Allocation units */
+ trans_class->transform_size =
+ GST_DEBUG_FUNCPTR (gst_tensordec_transform_size);
}
/**
* initialize instance structure
*/
static void
-gst_tensordec_init (GstTensorDec * filter)
+gst_tensordec_init (GstTensorDec * self)
{
- filter->silent = TRUE;
- filter->Configured = FALSE;
- filter->negotiated = FALSE;
- filter->addPadding = FALSE;
+ self->silent = DEFAULT_SILENT;
+ self->configured = FALSE;
+ self->negotiated = FALSE;
+ self->add_padding = FALSE;
+ gst_tensor_config_init (&self->tensor_config);
}
/**
- * @brief @todo fill this in
+ * @brief Set property (GObject vmethod)
*/
static void
gst_tensordec_set_property (GObject * object, guint prop_id,
const GValue * value, GParamSpec * pspec)
{
- GstTensorDec *filter = GST_TENSORDEC (object);
+ GstTensorDec *self = GST_TENSORDEC (object);
switch (prop_id) {
case PROP_SILENT:
- filter->silent = g_value_get_boolean (value);
+ self->silent = g_value_get_boolean (value);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
}
/**
- * @brief @todo fill this in
+ * @brief Get property (GObject vmethod)
*/
static void
gst_tensordec_get_property (GObject * object, guint prop_id,
GValue * value, GParamSpec * pspec)
{
- GstTensorDec *filter = GST_TENSORDEC (object);
+ GstTensorDec *self = GST_TENSORDEC (object);
switch (prop_id) {
case PROP_SILENT:
- g_value_set_boolean (value, filter->silent);
+ g_value_set_boolean (value, self->silent);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
}
/**
- * @brief Return 1 if we need to add stride per row from the stream data
- */
-static int
-add_stride_padding_per_row (const gchar * format, int width)
-{
- /** @todo The actual list is much longer. fill them (read https://gstreamer.freedesktop.org/documentation/design/mediatype-video-raw.html ) */
- if ((!g_strcmp0 (format, "RGB") || !g_strcmp0 (format, "BGR")
- || !g_strcmp0 (format, "I420")) && (width % 4))
- return 1;
- return 0;
-}
-
-
-/******************************************************************
- * GstElement vmethod implementations
- */
-
-#define return_false_if_fail(x) \
- ret = (x); \
- if (!ret) \
- return FALSE; \
- ;
-static const gchar format_RGB[] = "RGB";
-static const gchar format_BGRx[] = "BGRx";
-static const gchar interlace_progressive[] = "progressive";
-
-/**
* @brief Configure tensor metadata from sink caps
*/
static gboolean
-gst_tensordec_configure (const GstCaps * caps, GstTensorDec * filter)
+gst_tensordec_configure (GstTensorDec * self, const GstCaps * caps)
{
GstStructure *structure;
- tensor_type type;
- gint framerate_numerator;
- gint framerate_denominator;
- tensor_dim dimension;
- gint dim;
- gboolean ret;
- const gchar *format;
- const gchar *interlace;
- const gchar *type_str;
-
- /* This caps is coming from tensor */
- structure = gst_caps_get_structure (caps, 0);
- return_false_if_fail (gst_structure_get_int (structure, "dim1", &dim));
- dimension[0] = (uint32_t) dim;
- /** @todo Need to support orther media format (RGB, BRG, YUV,.. etc.). And should support Audio as well. */
- filter->output_media_type = _NNS_VIDEO;
- if (dimension[0] == 3 || dimension[0] == 4) {
- filter->format = dimension[0];
- if (dimension[0] == 3)
- format = format_RGB;
- else
- format = format_BGRx;
- } else {
- return FALSE;
- }
+ GstTensorConfig config;
- return_false_if_fail (gst_structure_get_int (structure, "dim2", &dim));
- dimension[1] = (uint32_t) dim;
- return_false_if_fail (gst_structure_get_int (structure, "dim3", &dim));
- dimension[2] = (uint32_t) dim;
- return_false_if_fail (gst_structure_get_fraction (structure, "framerate",
- &framerate_numerator, &framerate_denominator));
+ /** This caps is coming from tensor */
+ structure = gst_caps_get_structure (caps, 0);
- type_str = gst_structure_get_string (structure, "type");
- if (!g_strcmp0 (type_str, "uint8")) {
- type = _NNS_UINT8;
- } else {
+ if (!gst_tensor_config_from_structure (&config, structure)) {
+ err_print ("Cannot configure tensor from structure");
return FALSE;
}
-
- /* Emit Warning if RSTRIDE = RU4 (3BPP) && Width % 4 > 0 */
- /** @todo: Add more conditions! */
- if (add_stride_padding_per_row (format, dimension[1])) {
- g_print
- (" Width(dim2) is not divisible with 4. The performance won't be good; one more memcpy is added.\n");
- dlog_print (DLOG_WARN, "nnstreamer",
- "Input video width is not divisible with 4. The performance will not be good.");
- filter->addPadding = TRUE;
- }
-
- if (filter->Configured == TRUE) {
- /* It has been already configured. Check if they are consistent */
- if (dimension[1] == filter->dimension[1] &&
- dimension[2] == filter->dimension[2] &&
- type == filter->type &&
- framerate_numerator == filter->framerate_numerator &&
- framerate_denominator == filter->framerate_denominator) {
- return TRUE;
- }
- err_print (" Something's wrong. The tensor metadata is inconsistent.\n");
+ if (!gst_tensor_config_validate (&config)) {
+ err_print ("Not configured yet");
return FALSE;
}
- filter->type = type;
- filter->framerate_numerator = framerate_numerator;
- filter->framerate_denominator = framerate_denominator;
- filter->dimension[0] = dimension[0];
- filter->dimension[1] = dimension[1];
- filter->dimension[2] = dimension[2];
- filter->dimension[3] = 1;
-
- filter->Configured = TRUE;
-
- /** @todo Need to specify of video mode */
- filter->views = 1;
- interlace = interlace_progressive;
- if (!g_strcmp0 (interlace, "progressive")) {
- filter->mode = _VIDEO_PROGRESSIVE;
- } else {
- /* dose not support others for now */
+ if (self->configured && !gst_tensordec_check_consistency (self, &config)) {
+ err_print ("Mismatched to old metadata");
return FALSE;
}
- /** @todo Support other types */
- filter->output_media_type = _NNS_VIDEO;
- return TRUE;
-}
+ switch (config.tensor_media_type) {
+ case _NNS_VIDEO:
+ {
+ GstTensorVideoInfo v_info;
+ if (gst_tensor_video_info_from_config (&v_info, &config) &&
+ gst_tensor_video_stride_padding_per_row (v_info.format, v_info.w)) {
+ self->add_padding = TRUE;
+ }
-/**
- * @brief entry point to initialize the plug-in
- * initialize the plug-in itself
- * register the element factories and other features
- */
-static gboolean
-tensordec_init (GstPlugin * tensordec)
-{
- /* debug category for fltering log messages
- *
- * exchange the string 'Template tensordec' with your description
- */
- GST_DEBUG_CATEGORY_INIT (gst_tensordec_debug, "tensordec",
- 0, "Template tensordec");
+ break;
+ }
+ case _NNS_AUDIO:
+ break;
+ default:
+ err_print ("Unsupported type %d\n", config.tensor_media_type);
+ return FALSE;
+ }
- return gst_element_register (tensordec, "tensordec", GST_RANK_NONE,
- GST_TYPE_TENSORDEC);
+ self->tensor_config = config;
+ self->configured = TRUE;
+ return TRUE;
}
-/* PACKAGE: this is usually set by autotools depending on some _INIT macro
- * in configure.ac and then written into and defined in config.h, but we can
- * just set it ourselves here in case someone doesn't use autotools to
- * compile this code. GST_PLUGIN_DEFINE needs PACKAGE to be defined.
- */
-#ifndef PACKAGE
-#define PACKAGE "tensordec"
-#endif
-
-/* gstreamer looks for this structure to register tensordecs
- *
- * exchange the string 'Template tensordec' with your tensordec description
- */
-GST_PLUGIN_DEFINE (GST_VERSION_MAJOR,
- GST_VERSION_MINOR,
- tensordec,
- "tensordec",
- tensordec_init, VERSION, "LGPL", "GStreamer", "http://gstreamer.net/");
-
/**
- * @brief @todo fill this in
+ * @brief copies sink pad buffer to src pad buffer (internal static function)
+ * @param self "this" pointer
+ * @param inbuf sink pad buffer
+ * @param outbuf src pad buffer
+ * @return GST_FLOW_OK if ok. other values represents error
*/
static GstFlowReturn
-gst_t2c_transform (GstTensorDec * filter, GstBuffer * inbuf, GstBuffer * outbuf)
+gst_tensordec_copy_buffer (GstTensorDec * self,
+ GstBuffer * inbuf, GstBuffer * outbuf)
{
GstMapInfo inInfo, outInfo;
uint8_t *inptr, *outptr;
+ GstTensorConfig *config;
unsigned int row, d0;
unsigned int dest_idx = 0, src_idx = 0;
+ size_t size, offset, size_out;
+
+ g_assert (self->configured);
- size_t size = filter->dimension[0] * filter->dimension[1];
- size_t offset = size;
+ config = &self->tensor_config;
+
+ /** flag add_padding only for video */
+ g_assert (self->add_padding);
+ g_assert (config->tensor_media_type == _NNS_VIDEO);
+
+ size = offset = config->dimension[0] * config->dimension[1];
if (offset % 4)
offset += 4 - (offset % 4);
- size_t size_out = offset * filter->dimension[2] * filter->dimension[3];
+ size_out = offset * config->dimension[2] * config->dimension[3];
+
+ if (gst_buffer_get_size (outbuf) < size_out) {
+ gst_buffer_set_size (outbuf, size_out);
+ }
- g_assert (outbuf);
- if (filter->addPadding) {
- if (gst_buffer_get_size (outbuf) < size_out)
- gst_buffer_set_size (outbuf, size_out);
+ gst_buffer_map (inbuf, &inInfo, GST_MAP_READ);
+ gst_buffer_map (outbuf, &outInfo, GST_MAP_WRITE);
- gst_buffer_map (inbuf, &inInfo, GST_MAP_READ);
- gst_buffer_map (outbuf, &outInfo, GST_MAP_WRITE);
- inptr = inInfo.data;
- outptr = outInfo.data;
+ inptr = inInfo.data;
+ outptr = outInfo.data;
- for (d0 = 0; d0 < filter->dimension[3]; d0++) {
- g_assert (d0 == 0);
- for (row = 0; row < filter->dimension[2]; row++) {
- memcpy (outptr + dest_idx, inptr + src_idx, size);
- dest_idx += offset;
- src_idx += size;
- }
+ for (d0 = 0; d0 < config->dimension[3]; d0++) {
+ g_assert (d0 == 0);
+ for (row = 0; row < config->dimension[2]; row++) {
+ memcpy (outptr + dest_idx, inptr + src_idx, size);
+ dest_idx += offset;
+ src_idx += size;
}
- gst_buffer_unmap (inbuf, &inInfo);
- gst_buffer_unmap (outbuf, &outInfo);
}
+ gst_buffer_unmap (inbuf, &inInfo);
+ gst_buffer_unmap (outbuf, &outInfo);
+
return GST_FLOW_OK;
}
/**
- * @brief @todo fill this in
+ * @brief non-ip transform. required vmethod for BaseTransform class.
*/
static GstFlowReturn
gst_tensordec_transform (GstBaseTransform * trans,
GstBuffer * inbuf, GstBuffer * outbuf)
{
- GstTensorDec *filter = GST_TENSORDEC_CAST (trans);
+ GstTensorDec *self;
+ GstTensorConfig *config;
GstFlowReturn res;
- if (G_UNLIKELY (!filter->negotiated))
+
+ self = GST_TENSORDEC_CAST (trans);
+
+ if (G_UNLIKELY (!self->negotiated))
goto unknown_tensor;
- if (G_UNLIKELY (!filter->Configured))
+ if (G_UNLIKELY (!self->configured))
goto unknown_format;
- switch (filter->output_media_type) {
+ config = &self->tensor_config;
+
+ switch (config->tensor_media_type) {
case _NNS_VIDEO:
- res = gst_t2c_transform (filter, inbuf, outbuf);
- break;
- /* NOT SUPPORTED */
case _NNS_AUDIO:
+ res = gst_tensordec_copy_buffer (self, inbuf, outbuf);
+ break;
case _NNS_STRING:
default:
- err_print (" Unsupported Media Type (%d)\n", filter->output_media_type);
+ err_print ("Unsupported Media Type (%d)\n", config->tensor_media_type);
goto unknown_type;
}
return res;
unknown_format:
- GST_ELEMENT_ERROR (filter, CORE, NOT_IMPLEMENTED, (NULL), ("unknown format"));
+ GST_ELEMENT_ERROR (self, CORE, NOT_IMPLEMENTED, (NULL), ("unknown format"));
return GST_FLOW_NOT_NEGOTIATED;
unknown_tensor:
- GST_ELEMENT_ERROR (filter, CORE, NOT_IMPLEMENTED, (NULL),
+ GST_ELEMENT_ERROR (self, CORE, NOT_IMPLEMENTED, (NULL),
("unknown format for tensor"));
return GST_FLOW_NOT_NEGOTIATED;
unknown_type:
- GST_ELEMENT_ERROR (filter, CORE, NOT_IMPLEMENTED, (NULL),
+ GST_ELEMENT_ERROR (self, CORE, NOT_IMPLEMENTED, (NULL),
("not implemented type of media"));
return GST_FLOW_NOT_SUPPORTED;
}
/**
- * @brief @todo fill this in
+ * @brief in-place transform. required vmethod for BaseTransform class.
*/
static GstFlowReturn
gst_tensordec_transform_ip (GstBaseTransform * trans, GstBuffer * buf)
{
- GstTensorDec *filter = GST_TENSORDEC_CAST (trans);
+ GstTensorDec *self;
+ GstTensorConfig *config;
+
+ self = GST_TENSORDEC_CAST (trans);
- if (G_UNLIKELY (!filter->negotiated))
+ if (G_UNLIKELY (!self->negotiated))
goto unknown_format;
- if (G_UNLIKELY (!filter->Configured))
+ if (G_UNLIKELY (!self->configured))
goto unknown_tensor;
- switch (filter->output_media_type) {
+ config = &self->tensor_config;
+
+ switch (config->tensor_media_type) {
case _NNS_VIDEO:
- if (filter->addPadding == TRUE) {
- /** @todo Add Padding for x-raw */
+ if (self->add_padding) {
+ /**
+ * @todo Do we need to add padding for x-raw here?
+ */
}
break;
- /* NOT SUPPORTED */
case _NNS_AUDIO:
+ break;
case _NNS_STRING:
default:
- err_print (" Unsupported Media Type (%d)\n", filter->output_media_type);
+ err_print ("Unsupported Media Type (%d)\n", config->tensor_media_type);
goto unknown_type;
}
- /* DO NOTHING. THIS WORKS AS A PASSTHROUGH. We just remove metadata from video */
+ /** DO NOTHING. THIS WORKS AS A PASSTHROUGH. We just remove metadata from video */
return GST_FLOW_OK;
unknown_format:
- GST_ELEMENT_ERROR (filter, CORE, NOT_IMPLEMENTED, (NULL), ("unknown format"));
+ GST_ELEMENT_ERROR (self, CORE, NOT_IMPLEMENTED, (NULL), ("unknown format"));
return GST_FLOW_NOT_NEGOTIATED;
unknown_tensor:
- GST_ELEMENT_ERROR (filter, CORE, NOT_IMPLEMENTED, (NULL),
+ GST_ELEMENT_ERROR (self, CORE, NOT_IMPLEMENTED, (NULL),
("unknown format for tensor"));
return GST_FLOW_NOT_NEGOTIATED;
unknown_type:
- GST_ELEMENT_ERROR (filter, CORE, NOT_IMPLEMENTED, (NULL),
+ GST_ELEMENT_ERROR (self, CORE, NOT_IMPLEMENTED, (NULL),
("not implemented type of media"));
return GST_FLOW_NOT_SUPPORTED;
}
gst_tensordec_transform_caps (GstBaseTransform * trans,
GstPadDirection direction, GstCaps * caps, GstCaps * filter)
{
- GstCaps *tmp;
- gboolean ret;
- GstTensorDec bogusFilter = { 0 };
- bogusFilter.Configured = FALSE;
- GstTensorDec *obj = GST_TENSORDEC_CAST (trans);
+ GstTensorDec *self;
+ GstCaps *result;
- /* if direction is sink, check src (depending on sink's format, we could choose video or autiod. But currently only video/x-raw (RGB and RGBx) support. cap is for tensor. */
- if (direction == GST_PAD_SINK) {
- GstStructure *structure;
- gchar *str;
-
- ret = gst_tensordec_configure (caps, &bogusFilter);
-
- if (ret == FALSE) {
- GstStructure *structure = gst_caps_get_structure (caps, 0);
- gchar *str = gst_structure_to_string (structure);
- gchar str2[2048];
- gchar framerate[1024], width[1024], height[1024], colors[1024];
- /* const gchar *format; */
- int fn = -1, fd, w, h, c;
-
- if (TRUE == gst_structure_get_fraction (structure, "framerate", &fn, &fd))
- g_sprintf (framerate, "%d/%d", fn, fd);
- else
- g_sprintf (framerate, "[ 0/1, 2147483647/1 ]");
-
- if (TRUE == gst_structure_get_int (structure, "dim2", &w))
- g_sprintf (width, "%d", w);
- else
- g_sprintf (width, "[1, 65535]");
-
- if (TRUE == gst_structure_get_int (structure, "dim3", &h))
- g_sprintf (height, "%d", h);
- else
- g_sprintf (height, "[1, 65535]");
-
- if (TRUE == gst_structure_get_int (structure, "dim1", &c)) {
- if (c == 3)
- g_sprintf (colors, "%s", "RGB");
- else if (c == 4)
- g_sprintf (colors, "%s", "RGBx");
- } else {
- g_sprintf (colors, "{RGB, BGRx}");
- }
+ self = GST_TENSORDEC_CAST (trans);
- debug_print (!obj->silent, "Structure from caps = %s\n", str);
-
- g_sprintf (str2,
- "video/x-raw, "
- "format = (string)%s, "
- "views = (int)1, "
- "interlace-mode = (string)progressive, "
- "framerate = (fraction) %s, "
- "width = (int) %s, "
- "height = (int) %s", colors, framerate, width, height);
-
- tmp = gst_caps_from_string (str2);
- debug_print (!obj->silent, "Structure from caps to = %s\n", str2);
- g_free (str);
- /* If given caps are in range for width/height,
- we cannot configure tensor, however, we may return proper srcpad caps */
- /** @todo: see if the error is from ranging width/height before entering here */
- return tmp;
- }
- debug_print (!obj->silent, "transform_caps SINK specific\n");
-
- g_assert (bogusFilter.Configured == TRUE);
-
- /** @todo Need to support other format */
- gchar color[1024], interlace[1024];
- if (bogusFilter.format == 3)
- g_sprintf (color, "%s", "RGB");
- else if (bogusFilter.format == 4)
- g_sprintf (color, "%s", "BGRx");
- else
- g_sprintf (color, "%s", "{RGB, BGRx}");
- if (bogusFilter.mode == _VIDEO_PROGRESSIVE)
- g_sprintf (interlace, "progressive");
-
- tmp = gst_caps_new_simple ("video/x-raw",
- "width", G_TYPE_INT, bogusFilter.dimension[1],
- "height", G_TYPE_INT, bogusFilter.dimension[2],
- "format", G_TYPE_STRING, color,
- "views", G_TYPE_INT, bogusFilter.views,
- "interlace-mode", G_TYPE_STRING, interlace,
- "framerate", GST_TYPE_FRACTION, bogusFilter.framerate_numerator,
- bogusFilter.framerate_denominator, NULL);
-
- if (filter) {
- GstCaps *tmp2 =
- gst_caps_intersect_full (filter, tmp, GST_CAPS_INTERSECT_FIRST);
- gst_caps_unref (tmp);
- tmp = tmp2;
- }
- structure = gst_caps_get_structure (caps, 0);
- str = gst_structure_to_string (structure);
- debug_print (!obj->silent, "From = %s\n", str);
- g_free (str);
- structure = gst_caps_get_structure (tmp, 0);
- str = gst_structure_to_string (structure);
- debug_print (!obj->silent, "To = %s\n", str);
- g_free (str);
- GST_DEBUG_OBJECT (trans, "SINK transformed %" GST_PTR_FORMAT " into %"
- GST_PTR_FORMAT, caps, tmp);
- return tmp;
+ silent_debug_caps (caps, "from");
+ silent_debug_caps (filter, "filter");
+
+ /**
+ * If direction is sink, check src. Depending on sink's format, we could choose video or audio.
+ * Currently video/x-raw and audio/x-raw supported.
+ */
+ if (direction == GST_PAD_SINK) {
+ GstStructure *s = gst_caps_get_structure (caps, 0);
+ result = gst_tensordec_media_caps_from_structure (self, s);
} else if (direction == GST_PAD_SRC) {
- /* Construct possible GstCap (sinkpad) with src_factory */
- /** @todo This supports video only! */
- GstStaticCaps staticcap = GST_STATIC_CAPS (GST_TENSOR_CAP_DEFAULT);
- tmp = gst_static_caps_get (&staticcap);
+ /** caps from tensor */
+ GstStaticCaps raw_caps = GST_STATIC_CAPS (GST_TENSOR_CAP_DEFAULT);
+ result = gst_static_caps_get (&raw_caps);
+ } else {
+ silent_debug ("Direction = %d\n", direction);
+ g_assert (0);
+ return NULL;
+ }
+
+ if (filter) {
+ GstCaps *intersection;
+
+ intersection =
+ gst_caps_intersect_full (filter, result, GST_CAPS_INTERSECT_FIRST);
- return tmp;
+ gst_caps_unref (result);
+ result = intersection;
}
- /* Neither SRC/SINK? Impossible! */
- err_print ("Direction = %d\n", direction);
- GST_DEBUG_OBJECT (trans, "Error pad direction type. direction: %d",
- direction);
- g_assert (TRUE == FALSE);
- return NULL;
+
+ silent_debug_caps (result, "to");
+
+ GST_DEBUG_OBJECT (trans, "Direction[%d] transformed %" GST_PTR_FORMAT
+ " into %" GST_PTR_FORMAT, direction, caps, result);
+ return result;
}
/**
- * @brief @todo fill this in
+ * @brief fixate caps. required vmethod of BaseTransform
*/
static GstCaps *
gst_tensordec_fixate_caps (GstBaseTransform * trans,
GstPadDirection direction, GstCaps * caps, GstCaps * othercaps)
{
- GstCaps *supposed =
- gst_tensordec_transform_caps (trans, direction, caps, NULL);
+ GstTensorDec *self;
+ GstCaps *supposed;
GstCaps *result;
+ self = GST_TENSORDEC_CAST (trans);
+
+ silent_debug_caps (caps, "from caps");
+ silent_debug_caps (othercaps, "from othercaps");
+
GST_DEBUG_OBJECT (trans, "trying to fixate othercaps %" GST_PTR_FORMAT
" based on caps %" GST_PTR_FORMAT, othercaps, caps);
+ if (gst_tensordec_configure (self, caps)) {
+ supposed = gst_tensor_media_caps_from_config (&self->tensor_config);
+ } else {
+ GstStructure *s = gst_caps_get_structure (caps, 0);
+ supposed = gst_tensordec_media_caps_from_structure (self, s);
+ }
+
result = gst_caps_intersect (othercaps, supposed);
+ gst_caps_unref (supposed);
+
if (gst_caps_is_empty (result)) {
gst_caps_unref (result);
result = othercaps;
}
/**
- * @brief @todo fill this in
+ * @brief set caps. required vmethod of BaseTransform
*/
static gboolean
gst_tensordec_set_caps (GstBaseTransform * trans,
GstCaps * incaps, GstCaps * outcaps)
{
- GstTensorDec *filter = GST_TENSORDEC_CAST (trans);
- gboolean AddPadding = TRUE;
- int width, channel;
- const gchar *format;
- GstStructure *structure;
+ GstTensorDec *self;
- structure = gst_caps_get_structure (incaps, 0);
- gst_structure_get_int (structure, "dim2", &width);
- gst_structure_get_int (structure, "dim1", &channel);
+ self = GST_TENSORDEC_CAST (trans);
- if (channel == 3 || channel == 4) {
- if (channel == 3)
- format = format_RGB;
- else
- format = format_BGRx;
- } else {
- return FALSE;
+ silent_debug_caps (incaps, "from incaps");
+ silent_debug_caps (outcaps, "from outcaps");
+
+ /** compare and verify outcaps */
+ if (gst_tensordec_configure (self, incaps)) {
+ GstTensorConfig config;
+ GstStructure *s = gst_caps_get_structure (outcaps, 0);
+
+ if (gst_tensor_config_from_structure (&config, s) &&
+ gst_tensordec_check_consistency (self, &config)) {
+ self->negotiated = TRUE;
+ }
}
- if (add_stride_padding_per_row (format, width))
- AddPadding = FALSE;
+ gst_base_transform_set_in_place (trans, !self->add_padding);
+
+ return self->negotiated;
+}
+
+/**
+ * @brief Tell the framework the required size of buffer based on the info of the other side pad. optional vmethod of BaseTransform
+ *
+ * This is called when non-ip mode is used.
+ */
+static gboolean
+gst_tensordec_transform_size (GstBaseTransform * trans,
+ GstPadDirection direction, GstCaps * caps, gsize size,
+ GstCaps * othercaps, gsize * othersize)
+{
+ GstTensorDec *self;
+ GstTensorConfig *config;
+
+ self = GST_TENSORDEC_CAST (trans);
+
+ g_assert (self->configured);
- gst_base_transform_set_in_place (trans, AddPadding);
+ config = &self->tensor_config;
- filter->negotiated = gst_tensordec_configure (incaps, filter);
+ if (self->add_padding) {
+ gsize offset;
+
+ /** flag add_padding only for video */
+ g_assert (config->tensor_media_type == _NNS_VIDEO);
+
+ offset = config->dimension[0] * config->dimension[1];
+
+ if (offset % 4)
+ offset += 4 - (offset % 4);
+
+ *othersize = offset * config->dimension[2] * config->dimension[3];
+ } else {
+ *othersize = self->tensor_config.frame_size;
+ }
return TRUE;
}
+
+/**
+ * @brief entry point to initialize the plug-in
+ * initialize the plug-in itself
+ * register the element factories and other features
+ */
+static gboolean
+gst_tensordec_plugin_init (GstPlugin * plugin)
+{
+ /**
+ * debug category for fltering log messages
+ */
+ GST_DEBUG_CATEGORY_INIT (gst_tensordec_debug, "tensordec",
+ 0, "Element to convert tensor to media stream");
+
+ return gst_element_register (plugin, "tensordec", GST_RANK_NONE,
+ GST_TYPE_TENSORDEC);
+}
+
+/**
+ * PACKAGE: this is usually set by autotools depending on some _INIT macro
+ * in configure.ac and then written into and defined in config.h, but we can
+ * just set it ourselves here in case someone doesn't use autotools to
+ * compile this code. GST_PLUGIN_DEFINE needs PACKAGE to be defined.
+ */
+#ifndef PACKAGE
+#define PACKAGE "tensordec"
+#endif
+
+/**
+ * gstreamer looks for this structure to register tensordec
+ */
+GST_PLUGIN_DEFINE (GST_VERSION_MAJOR,
+ GST_VERSION_MINOR,
+ tensordec,
+ "Element to convert tensor to media stream",
+ gst_tensordec_plugin_init,
+ VERSION, "LGPL", "GStreamer", "http://gstreamer.net/");
projects / platform / upstream / nnstreamer.git / commitdiff